Intelligent efficient grinding robot workstation

Abstract

The invention discloses an intelligent efficient grinding robot workstation which comprises a detection unit, a grinding unit, a moving unit and a control system. The detection unit adopts a machine vision system and is provided with a dust cover, the three-dimensional shape of the defect is reconstructed by utilizing a workpiece image and a deep learning model, and various quantitative parameters are extracted; the grinding unit comprises a grinding robot, a force control device and a grinding tool; the moving unit drives the workpiece to be switched between the detection area and the grinding area through the sliding table module. The control system adaptively matches optimal polishing parameters from a preset process parameter library through an intelligent algorithm on the basis of the defect quantization parameters, controls the polishing robot to execute polishing, automatically rechecks after polishing is completed, and circularly repairs residual defects until the residual defects are qualified or are judged as defective products; the full-process closed-loop control of detection, matching, grinding, reinspection and repairing is achieved, and the problems that traditional grinding depends on artificial experience, dust interferes with detection, and the self-adaptive capacity is lacked are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of robotic grinding and processing technology, specifically to an intelligent and efficient grinding robot workstation. Background Technology

[0002] With the increasing demands for workpiece surface quality in the manufacturing industry, grinding processes are being used more and more widely in fields such as aerospace, automobile manufacturing, and mold processing. Currently, traditional grinding operations still mainly rely on manual operation or semi-automated equipment. Manual grinding suffers from problems such as high labor intensity, harsh working environment (dust, noise), unstable grinding quality, and low efficiency. Moreover, it requires a high level of technical experience from operators, making it difficult to meet the production needs of large-scale, high-precision, and diversified products.

[0003] To address the issue of manual labor, existing technologies have developed industrial robots for automated grinding, but these technologies still have the following shortcomings:

[0004] 1. Lack of intelligent perception capability: Traditional grinding robots are mostly in the "teach-and-playback" mode, which cannot perceive the location, type and severity of defects on the workpiece surface in real time. It is difficult to achieve adaptive grinding, resulting in over-grinding or under-grinding, or even damage to the workpiece substrate.

[0005] 2. Separation of inspection and grinding: In existing solutions, inspection and grinding are often separated into two independent workstations or equipment. The workpiece needs to be clamped and transferred multiple times, which is not only inefficient, but also easy to introduce positioning errors and affect grinding accuracy.

[0006] 3. Severe dust interference: At the polishing site, a large amount of metallic or non-metallic dust is present in the air. If the detection system (such as a vision camera) is directly exposed to the dust environment, the lens is easily contaminated, resulting in blurred images and detection failure, which affects subsequent polishing decisions.

[0007] 4. Lack of adaptive process adjustment capability: Different materials (such as cast iron, aluminum alloy, composite materials) and different defect types (such as burrs, pits, weld beads, oxide scale) have different requirements for grinding process parameters (pressure, angle, trajectory, number of times, etc.). Existing equipment is difficult to dynamically match the optimal process parameters based on the test results, and the level of intelligence is low.

[0008] 5. Lack of closed-loop feedback mechanism: Most existing polishing systems do not have the closed-loop control capability of "polishing-re-inspection-repair", and cannot perform real-time evaluation and targeted repair of the surface quality after polishing, which easily leads to defective products and increases rework costs.

[0009] Therefore, existing technologies cannot meet the demands of modern production for efficient, precise, and intelligent processing. There is an urgent need for an intelligent and efficient grinding robot workstation to improve the automation and intelligence level of grinding operations and ensure the consistency and efficiency of workpiece grinding quality. Summary of the Invention

[0010] In order to solve the many problems existing in the workpiece grinding technology mentioned in the background, the present invention provides an intelligent and efficient grinding robot workstation.

[0011] The above-mentioned objective of this application is achieved through the following technical solution:

[0012] A smart and efficient grinding robot workstation includes:

[0013] The inspection unit includes a machine vision system and a dust cover; the machine vision system uses a 3D camera or a 2D camera, which is fixedly installed above the inspection area to detect workpiece defect features; the dust cover is filled with positive pressure gas to prevent external dust from entering and affecting the operation of the machine vision system during the inspection process.

[0014] The grinding unit includes a high-rigidity, high-load grinding robot, a force control device, grinding tools, and consumables. The high-rigidity, high-load grinding robot is used to carry the grinding tools and execute the grinding trajectory. The force control device, with a flexible buffer function, is installed at the end of the grinding robot to precisely control the grinding pressure. The grinding tools and consumables are selected and adapted according to the workpiece material and defect type.

[0015] The moving unit includes a product clamping fixture and a moving slide module; the product clamping fixture is used to fix the workpiece; the moving slide module is connected to the product clamping fixture and is used to drive the product clamping fixture to switch between the inspection area and the grinding area.

[0016] The control system is communicatively connected to the detection unit, grinding unit, and moving unit, and is used to control the grinding actuator to perform grinding operations on the workpiece based on the quantitative parameters of the defect.

[0017] By adopting the above technical solution, the detection unit, grinding unit, and moving unit are integrated into the same workstation, realizing the integrated operation of automatic identification and precise grinding of workpiece surface defects. The machine vision system, combined with a positive pressure dust cover, can stably acquire high-precision image data in a clean environment, ensuring the accuracy and repeatability of defect identification. The grinding unit, through a high-rigidity robot combined with a flexible force control device, can ensure both rigidity and trajectory accuracy during the grinding process, and dynamically adjust the contact force according to the surface characteristics of the workpiece, avoiding damage to the workpiece from hard impacts and improving grinding quality and consistency. The moving unit drives the workpiece to quickly switch between the detection area and the grinding area. This system achieves a closed-loop process of inspection, grinding, and re-inspection, reducing manual intervention and improving operational efficiency. The control system uses intelligent algorithms to comprehensively analyze the defect quantification parameters obtained by the inspection unit, adaptively matching the optimal combination of process parameters from a preset process parameter library, and controlling the grinding unit to perform precise grinding operations according to the matched parameters. This forms a fully intelligent closed-loop control of the "inspection-analysis-grinding-re-inspection" process. This workstation effectively solves the problems of relying on manual experience, lack of real-time feedback, and dust interference detection in the traditional grinding process, improving grinding accuracy, adaptability, and automation level. It is suitable for intelligent processing needs of various materials and complex defect scenarios.

[0018] In a preferred embodiment, the present application may be further configured such that the machine vision system further includes a deep learning processing unit, which performs three-dimensional reconstruction of defects based on deep learning technology and extracts defect quantification parameters including defect type, location, area, height, volume and severity.

[0019] By adopting the above technical solutions, the machine vision system integrates a deep learning processing unit, which can intelligently analyze the acquired workpiece surface images. It can not only identify the type and location of defects, but also realize the three-dimensional reconstruction of defects and accurately calculate quantitative parameters such as the area, height, volume and severity of defects. This provides comprehensive and accurate data support for subsequent grinding processes, upgrading grinding operations from traditional experience-based judgment to data-driven precision control, and improving the ability to identify complex defects and the accuracy of process matching.

[0020] In a preferred embodiment, the dust cover may be further configured such that it has a compressed air inlet for introducing compressed air into the dust cover and maintaining positive pressure inside the cover.

[0021] By adopting the above technical solution, the dust cover, by setting up a compressed air inlet and maintaining positive pressure inside the cover, can continuously block external grinding dust from entering the working area of ​​the vision system during the inspection process, effectively preventing dust from adhering to the lens or interfering with image acquisition, ensuring that the machine vision system can still operate stably in harsh grinding environments, and guaranteeing the inspection accuracy and reliability of the machine vision system.

[0022] In a preferred embodiment, this application may be further configured to include an auxiliary functional unit, which includes a rack, a water circulation device, an automatic consumable replacement device, and a dust extraction device.

[0023] The rack is used to support and install the various functional modules;

[0024] The water circulation device is used for cooling during the grinding process;

[0025] The automatic consumable replacement device is used to replace the polishing consumables in real time according to the wear condition of the consumables.

[0026] The exhaust dust removal device is used to handle the dust generated during grinding.

[0027] By adopting the above technical solutions, the auxiliary functional units provide comprehensive guarantees for the efficient and stable operation of the workstation: the frame, as the overall support structure, ensures the integrated installation and collaborative operation of each functional module; the water circulation device continuously provides cooling during the grinding process, effectively reducing the temperature of the grinding area, preventing workpiece thermal deformation or consumable overheating failure, extending tool life, and improving grinding quality; the automatic consumable replacement device can monitor the wear status of consumables in real time and automatically complete the replacement according to preset conditions, avoiding the decline in grinding effect or work interruption caused by consumable wear, and improving the continuous operation capability and automation level of the equipment; the exhaust dust removal device simultaneously handles the dust generated during grinding, which not only improves the working environment and reduces the harm of dust to equipment and personnel, but also prevents secondary dust contamination of the workpiece surface or interference with the detection unit, further ensuring grinding accuracy and control system stability; the integration of the auxiliary functional units enables the workstation to have the ability to operate continuously for a long time, under high load, and with high precision, realizing the intelligent, automated, and environmentally friendly grinding process.

[0028] The second objective of this invention is achieved through the following technical solution:

[0029] The intelligent and efficient grinding robot workstation employs an intelligent grinding control method, which includes the following steps:

[0030] The workpiece is scanned and photographed by a machine vision system fixedly installed above the inspection area. Based on deep learning technology, the defects are reconstructed in three dimensions, and quantitative parameters of defects, including defect type, location, area, height, volume and severity, are extracted.

[0031] Establish a grinding process parameter library covering different materials and defect types. The process parameters include grinding angle, pressure, trajectory, number of passes, moving speed, feed rate, and dry / wet grinding modes.

[0032] By comprehensively analyzing the defect quantification parameters through intelligent algorithms, the optimal combination of process parameters is adaptively selected from the process parameter library;

[0033] The workpiece is switched from the inspection area to the grinding area by moving the product clamping fixture through the sliding table module;

[0034] The high-rigidity, high-load grinding robot performs grinding operations under the flexible buffer control of the force control device, based on adaptively matched process parameters and planned trajectories.

[0035] After grinding, the workpiece is moved back to the inspection area for re-inspection. If it fails, targeted repair grinding is carried out according to the type, location and severity of the unfinished defects until it passes or is judged as a defective product.

[0036] By adopting the above technical solutions, intelligent perception, process decision-making, and flexible execution are deeply integrated to construct a complete closed-loop grinding control process. The machine vision system, based on deep learning, performs three-dimensional reconstruction and quantitative analysis of workpiece surface defects, transforming traditionally difficult-to-describe defect features into calculable process parameter inputs, providing a precise data foundation for subsequent process matching. By establishing a process parameter library covering multiple materials and defect types, and combining intelligent algorithms to comprehensively analyze defect quantification parameters, the control system can adaptively match the optimal combination of grinding angle, pressure, trajectory, and other parameters, improving its adaptability to different workpieces and complex defect scenarios. During execution, the moving unit automatically switches workpieces. Upon reaching the grinding area, the high-rigidity, high-load robot precisely executes the grinding operation under the flexible buffer control of the force control device. This ensures both the rigidity and accuracy of the grinding trajectory while avoiding hard damage to the workpiece through constant pressure control. After grinding, the robot automatically moves back to the inspection area for re-inspection. If residual defects are found, targeted repair grinding is performed based on updated quantitative parameters, forming an intelligent closed loop of "inspection—matching—grinding—re-inspection—repair" until the workpiece is qualified or determined to be defective. This control method achieves full automation and intelligence in the grinding process, which not only significantly improves grinding accuracy and consistency but also effectively reduces manual intervention and material waste. It is suitable for grinding complex workpieces with high precision, multiple varieties, and small batches.

[0037] In a preferred embodiment, this application can be further configured such that: the step of scanning and photographing the workpiece using a machine vision system fixedly installed above the detection area, and realizing three-dimensional reconstruction of defects based on deep learning technology, specifically includes:

[0038] Compressed air is introduced into the dust cover to maintain positive pressure inside the cover and prevent external grinding dust from entering and affecting the machine vision operation.

[0039] The workpiece is scanned and photographed by a 3D or 2D camera that is fixedly installed above the inspection area to obtain image data of the workpiece surface.

[0040] Image data is input into a deep learning model to identify defect areas and classify defect types.

[0041] The identified defect areas are reconstructed in three dimensions, and the geometric feature parameters of the defects are calculated, including the defect contour, projected area, maximum height, volume and curvature change.

[0042] The severity level of the defect is determined by comparing the geometric feature parameters with a preset threshold.

[0043] The defect type, location coordinates, geometric feature parameters, and severity level are combined to form a complete set of defect quantification parameters.

[0044] By adopting the above technical solution, and by continuously introducing compressed air into the dust cover and maintaining a positive pressure environment, the interference of dust generated during the grinding process on the machine vision system is effectively blocked, ensuring the clarity and stability of image acquisition and providing a reliable data foundation for subsequent analysis. Then, by scanning and photographing the workpiece with a 3D or 2D camera to obtain surface image data and inputting it into a deep learning model, the machine vision system can automatically identify defect areas and accurately classify defect types, overcoming the limitations of traditional manual visual inspection or simple image processing in identifying complex defects. By performing three-dimensional reconstruction on the identified defect areas, not only are basic geometric features such as the contour, projected area, and maximum height of the defects calculated, but also refined parameters such as volume and curvature changes are extracted, elevating the representation of defects from a two-dimensional plane to a three-dimensional spatial dimension, greatly enriching the dimension and accuracy of defect information. Finally, by comparing the geometric feature parameters with preset thresholds, the severity level of the defects is scientifically determined, and the defect type, location coordinates, geometric feature parameters, and severity level are integrated into a complete set of quantitative parameters, providing comprehensive and accurate input data for the adaptive matching of subsequent grinding processes.

[0045] In a preferred embodiment, this application can be further configured such that the step of comprehensively analyzing defect quantification parameters using an intelligent algorithm and adaptively selecting the optimal combination of process parameters from the process parameter library includes:

[0046] Construct a process parameter library that covers the correspondence between different workpiece materials, different defect types and grinding process parameters. The process parameters include grinding angle, pressure, trajectory, number of passes, moving speed, feed rate and dry and wet grinding modes.

[0047] The extracted defect type, location, area, height, volume, and severity are used as input features;

[0048] The input features are comprehensively analyzed by intelligent algorithms to calculate the matching degree between the current defect features and each case in the process parameter library;

[0049] The process parameter combination with the highest similarity is selected from the process parameter library as the basic parameter;

[0050] The basic parameters are adjusted linearly or nonlinearly based on the severity of the defects to generate the optimal combination of process parameters.

[0051] By adopting the above technical solution, and establishing a process parameter library covering different workpiece materials and defect types, the system associates diverse process parameters such as grinding angle, pressure, trajectory, number of passes, moving speed, feed rate, and dry / wet modes with specific working conditions, forming a reusable process knowledge base. This lays the data foundation for subsequent intelligent matching. After defect detection, the control system uses the extracted quantitative parameters such as defect type, location, area, height, volume, and severity as input features. Through intelligent algorithms, it calculates and comprehensively analyzes the similarity between the current defect features and various cases in the process parameter library. This allows for the rapid matching of the basic parameter combination closest to the current working condition from historical process data, achieving effective reuse and transfer of process knowledge. Based on this, the system further adjusts the basic parameters linearly or nonlinearly according to the defect severity, generating the optimal process parameter combination for the current workpiece. This ensures that process decisions not only rely on historical experience but also can be dynamically optimized and precisely adapted based on the actual defect situation. This improves the process adaptability to workpieces of different materials and defect types, ensuring the accuracy, efficiency, and consistency of the grinding process.

[0052] In a preferred embodiment, this application can be further configured such that the high-rigidity, high-load grinding robot performs grinding operations under the flexible buffer control of a force control device, based on adaptively matched process parameters and a planned trajectory, including the following steps:

[0053] Based on the location coordinates and outline of the defect, a local fixed-point grinding trajectory is planned for the defect area, so that the grinding tool only covers the defect area and its neighborhood, avoiding full grinding of non-defect areas.

[0054] The matching grinding pressure value is sent to the force control device, so that the force control device maintains a constant output force during the grinding process, realizing flexible buffer control and avoiding hard collisions that damage the workpiece.

[0055] The force control device provides real-time feedback of contact force data. The high-rigidity, high-load grinding robot dynamically adjusts its feed speed and attitude angle based on the deviation between the real-time contact force and the target pressure, thus precisely controlling the grinding pressure.

[0056] Simultaneously start the water circulation device and the dust extraction and removal device for cooling and dust treatment;

[0057] The automatic consumable replacement device replaces the polishing consumables in real time according to the wear condition of the consumables or the preset replacement cycle to ensure the polishing effect.

[0058] By adopting the above technical solution, and planning a local fixed-point grinding trajectory based on the defect location coordinates and contour, the grinding tool only acts on the defect area and its neighborhood, avoiding the ineffective processing and material loss of non-defect areas by traditional full-area grinding, thus improving grinding efficiency and workpiece surface integrity. Regarding pressure control, after the matched grinding pressure value is sent to the force control device, the device maintains a constant output force during grinding, achieving flexible buffer control and effectively avoiding hard damage to the workpiece caused by rigid contact. The force control device provides real-time feedback of contact force data, and the high-rigidity, high-load grinding robot dynamically adjusts the feed speed and attitude angle according to the deviation between the actual contact force and the target pressure, forming a closed-loop pressure control to ensure that the pressure remains stable within the preset range during grinding, improving the efficiency of grinding. The system effectively adapts to complex surfaces and irregularly shaped workpieces, ensuring consistent grinding performance. In terms of auxiliary support, a synchronized water circulation system and dust extraction system continuously cool and treat dust during grinding, effectively reducing the temperature of the heat-affected zone, preventing workpiece thermal deformation and consumable overheating failure, while also improving the working environment and avoiding secondary dust pollution. An automatic consumable replacement system replaces grinding consumables in real time based on wear or a preset cycle, ensuring the grinding tools are always in optimal working condition and preventing a decline in grinding results or work interruptions due to consumable wear. Through the synergy of trajectory planning, constant pressure control, dynamic adjustment, and auxiliary support, a highly efficient, precise, and autonomous grinding execution system is constructed, improving the consistency of grinding quality, the continuous operation capability of the equipment, and its level of intelligence.

[0059] In a preferred embodiment, this application can be further configured as follows: after grinding, the workpiece is moved back to the inspection area for re-inspection; if it fails, targeted repair grinding is performed.

[0060] After grinding, the workpiece is moved from the grinding area back to the inspection area using the movable slide module;

[0061] Under the condition of maintaining positive pressure of the dust cover, the polished area is scanned and inspected again by the machine vision system fixedly installed above the inspection area to obtain the surface morphology data after polishing.

[0062] The surface morphology data obtained from the re-inspection is compared with the preset acceptance standard to determine whether the defects have been completely removed;

[0063] If a residual defect is detected, extract the quantitative parameters of the residual defect, including the updated location, area, height, and severity.

[0064] Based on the residual defect quantification parameters, the repair grinding process parameters are re-matched from the process parameter library using an intelligent algorithm;

[0065] The workpiece is moved back to the grinding area, and only the areas with residual defects are ground for targeted repair grinding.

[0066] Repeat the re-inspection and repair steps until the workpiece is qualified or the preset maximum number of grinding times is reached;

[0067] If the maximum number of polishing cycles is exceeded or the severity of the defect exceeds the repairable range, the product is judged as defective and the defect data and polishing history are recorded.

[0068] By adopting the above technical solution, a complete closed-loop quality control system was constructed after grinding, realizing data-driven and autonomous decision-making throughout the entire process from grinding to final judgment. After grinding, the workpiece automatically returns to the inspection area via a moving slide module. Under the positive pressure protection of the dust cover, the machine vision system re-scans and inspects the ground area, acquiring the surface morphology data after grinding and comparing it with preset acceptance standards. This achieves an objective and quantitative assessment of grinding quality, avoiding the subjectivity and uncertainty of manual visual inspection. If residual defects are detected, the machine vision system can accurately extract the updated position, area, height, and severity. The system uses a high-precision quantification parameter and an intelligent algorithm to re-match a targeted repair grinding process from the process parameter library. This eliminates reliance on experience-based judgment and allows for precise adaptation based on the actual residual condition. The workpiece is then moved back to the grinding area for targeted repair grinding of only the remaining defective areas, avoiding damage and unnecessary consumption of already qualified areas due to repeated processing. The entire process can be repeated until the workpiece is qualified or reaches the preset maximum number of grinding cycles. If the workpiece exceeds the repairable range, the machine vision system identifies it as a defective product and records the defect data and grinding history, providing data support for subsequent quality traceability and process optimization.

[0069] In summary, this application includes at least one of the following beneficial technical effects:

[0070] 1. By integrating the detection unit, grinding unit, and moving unit into the same workstation and adopting a closed-loop control process of "detection-matching-grinding-re-inspection-repair," the entire process of identifying and repairing workpiece surface defects is automated. The machine vision system stably acquires high-precision image data under the protection of a positive pressure dust cover, and combines deep learning to achieve three-dimensional reconstruction and quantitative analysis of defects, providing accurate input for the grinding process. After grinding, automatic re-inspection is performed, and fixed-point repair is carried out according to the residual defects, ensuring that each grinding meets the preset standard. This effectively avoids the problems of insufficient or excessive grinding in traditional open-loop control, and significantly improves the consistency and pass rate of workpiece surface quality.

[0071] 2. By establishing a process parameter library covering multiple materials and defect types, and combining intelligent algorithms to perform similarity calculation and dynamic adjustment of defect quantification parameters, the system can adaptively match the optimal grinding angle, pressure, trajectory, number of passes, and other process parameters according to the actual situation of the workpiece. Regardless of the workpiece material or the complex and varied defect types, the system can quickly generate targeted process solutions, eliminating the reliance on manual experience in traditional grinding and improving the process adaptability and response speed for small batches, multiple varieties, and high-precision workpieces.

[0072] 3. The high-rigidity, high-load grinding robot, combined with a force control device with flexible buffering function, maintains constant pressure output during the grinding process and dynamically adjusts the feed speed and attitude angle based on real-time contact force feedback. This achieves precise constant-pressure grinding of complex curved surfaces, avoiding damage to the workpiece caused by hard impacts. The synchronously operating water circulation device, dust extraction device, and automatic consumable replacement device provide cooling, dust removal, and consumable replacement guarantees, effectively reducing the heat impact, improving the working environment, and ensuring that the tools are always in optimal condition. This enables the workstation to operate continuously for long periods, under high loads, and with high precision, thereby improving production efficiency and equipment utilization.

[0073] 4. During the grinding process, the system fully records the defect quantification parameters, matching process parameters, grinding execution data, and re-inspection results for each inspection, forming a quality data chain for the entire process. When a workpiece is judged to be defective, the system saves complete defect data and grinding history, providing reliable data support for subsequent quality traceability, process analysis, and parameter optimization. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the overall structure of the intelligent and efficient grinding robot workstation of this application;

[0075] Figure 2 This is a cross-sectional view of the intelligent and efficient grinding robot workstation of this application;

[0076] Figure 3 This is a flowchart illustrating the system architecture of the intelligent and efficient grinding robot workstation used in this application.

[0077] Figure 4 This is a flowchart illustrating an implementation of the intelligent and efficient grinding robot workstation of this application.

[0078] The components include: 1. Detection unit; 10. Machine vision system; 11. Dust cover; 2. Grinding unit; 20. Grinding robot; 21. Force control device; 22. Grinding tool; 3. Moving unit; 30. Product clamping fixture; 31. Moving slide module. Detailed Implementation

[0079] The following is in conjunction with the appendix Figure 1-4This application will be described in further detail.

[0080] In one embodiment, such as Figure 1-3 As shown, this application discloses an intelligent and efficient grinding robot workstation 20, which specifically includes:

[0081] The detection unit 1 includes a machine vision system 10 and a dust cover 11. The machine vision system 10 uses a 3D camera or a 2D camera and is fixedly installed above the detection area to detect the defect features of the workpiece. The dust cover 11 is filled with positive pressure gas to prevent external dust from entering and affecting the operation of the machine vision system 10 during the detection process.

[0082] Grinding unit 2 includes a high-rigidity, high-load grinding robot 20, a force control device 21, grinding tools 22, and consumables. The high-rigidity, high-load grinding robot 20 carries the grinding tools 22 and executes the grinding trajectory. The force control device 21, with a flexible buffer function, is installed at the end of the grinding robot 20 and is used to precisely control the grinding pressure. The grinding tools 22 and consumables are selected and adapted according to the workpiece material and defect type.

[0083] The moving unit 3 includes a product clamping fixture 30 and a moving slide module 31; the product clamping fixture 30 is used to fix the workpiece; the moving slide module 31 is connected to the product clamping fixture 30 and is used to drive the product clamping fixture 30 to switch between the inspection area and the grinding area.

[0084] In this embodiment, the detection unit 1 scans and captures high-precision image data of the workpiece surface using a 3D or 2D camera fixedly installed above the detection area. The dust cover 11 is continuously filled with positive pressure gas, creating a micro-positive pressure environment in the detection area. This effectively prevents dust generated during the grinding process from entering the working area of ​​the vision system, ensuring the clarity and stability of the image acquisition and providing a reliable data foundation for subsequent defect identification. After analysis by the deep learning processing unit, the image data acquired by the machine vision system 10 can accurately identify defect areas and extract quantitative parameters such as defect type, location, area, height, volume, and severity. This allows for precise assessment of workpiece surface quality. The moving unit 3, via the moving slide module 31, drives the product clamping fixture 30, quickly switching the inspected workpiece from the inspection area to the grinding area, achieving seamless integration of inspection and grinding processes. In the grinding unit 2, the high-rigidity, high-load grinding robot 20 adaptively matches optimal grinding angle, pressure, trajectory, number of passes, moving speed, feed rate, and wet / dry mode from the process parameter library based on the defect quantification parameters output by the inspection unit 1, and plans a localized grinding trajectory for the defect area. The force control device 21, installed at the end of the grinding robot 20, provides flexible buffering. Yes, during the grinding process, the robot maintains a constant output force based on the matched pressure value, achieving flexible contact control and avoiding damage to the workpiece from rigid impacts. Simultaneously, the force control device 21 provides real-time feedback of contact force data, and the robot dynamically adjusts its feed speed and attitude angle based on the deviation between the actual contact force and the target pressure, forming a closed-loop pressure control. This ensures that the pressure remains stable within the preset range during grinding, improving its adaptability to complex curved surfaces and irregularly shaped workpieces. After grinding, the moving unit 3 moves the workpiece back to the inspection area for re-inspection. Under the positive pressure protection of the dust cover 11, the machine vision system 10 re-scans and inspects the ground area to obtain... The surface morphology data after polishing is compared with the preset acceptance standard. If there are residual defects, the system extracts the updated defect quantification parameters, rematches the repair polishing process parameters, and performs targeted repair polishing on the residual defect area. The process can be repeated until the workpiece is qualified or the preset maximum number of polishing times is reached. If it exceeds the repairable range, it is judged as a defective product and complete defect data and polishing history are recorded. The system realizes full-process automation and intelligence from defect detection, process matching, precision polishing to quality re-inspection, effectively solving the problems of relying on manual experience, lack of real-time feedback, and dust interference detection in the traditional polishing process.

[0085] In one embodiment, the machine vision system 10 further includes a deep learning processing unit, which performs three-dimensional reconstruction of defects based on deep learning technology and extracts defect quantification parameters including defect type, location, area, height, volume and severity.

[0086] In this embodiment, the deep learning processing unit of the machine vision system 10 is the core module for improving the accuracy and intelligence level of defect detection. After the workpiece is positioned in the detection area, a 3D camera or a 2D camera performs high-resolution scanning and imaging of the workpiece surface to obtain complete surface image data. This image data is transmitted to the deep learning processing unit in real time, and after preprocessing, it is input into the pre-trained deep learning model. The deep learning model first performs semantic segmentation and target detection on the image, accurately identifies the defect area and classifies its type, such as common surface defects like scratches, pits, pores, and burrs. Then, it further reconstructs the three-dimensional shape of the identified defect area using multi-view imaging or structured light principles, thereby calculating a series of accurate geometric figures. The system incorporates various characteristic parameters, including the defect's contour boundary, projected area, maximum height, volume, and surface curvature variation. These parameters not only describe the defect's distribution on a two-dimensional plane but also comprehensively characterize its geometric shape and severity in three-dimensional space. The deep learning processing unit compares the calculated geometric characteristic parameters with preset defect severity grading thresholds to automatically determine the severity level of each defect, such as minor, moderate, or severe. Finally, the system integrates the defect type, its precise location coordinates in the workpiece coordinate system, various geometric characteristic parameters, and severity level to form a complete set of defect quantification parameters. This parameter set serves as input data for subsequent grinding process decisions, providing a comprehensive and accurate basis for intelligent algorithms to match optimal process parameters.

[0087] In one embodiment, the dust cover 11 is provided with a compressed air inlet for introducing compressed air into the dust cover 11 and maintaining positive pressure inside the cover.

[0088] In this embodiment, the dust cover 11 is positioned above the detection area, physically isolating the machine vision system 10 from the external environment and providing comprehensive protection for the vision system's lens and sensors. The dust cover 11 is equipped with a compressed air inlet connected to an external compressed air source, through which clean compressed air is continuously supplied to the interior of the dust cover 11 via a pipeline. During actual operation, compressed air continuously enters the dust cover 11 at a set flow rate and pressure, causing the internal air pressure to be slightly higher than the external atmospheric pressure, creating a stable micro-positive pressure environment. This positive pressure difference effectively prevents suspended dust, metal shavings, and other fine particles generated in the grinding area from entering the interior of the dust cover 11 through gaps, avoiding dust adhering to the camera lens, sensor surface, or drifting in the optical path, thereby ensuring the clarity and stability of image acquisition. Furthermore, the continuously supplied compressed air creates a weak airflow within the cover, helping to remove the heat generated during the vision system's operation, thus providing some heat dissipation and extending the equipment's lifespan.

[0089] In one embodiment, an auxiliary function unit is also included, which includes a frame, a water circulation device, an automatic consumable replacement device, and a dust extraction device.

[0090] The rack is used to support and install the various functional modules;

[0091] The water circulation device is used for cooling during the grinding process;

[0092] The automatic consumable replacement device is used to replace the polishing consumables in real time according to the wear condition of the consumables.

[0093] The exhaust dust removal device is used to handle the dust generated during grinding.

[0094] In this embodiment, the frame, as the basic support structure of the entire workstation, is made of high-strength profiles welded or assembled, and has excellent rigidity and stability. The detection unit 1, grinding unit 2, moving unit 3 and various auxiliary devices are all reasonably arranged and fixedly installed on the frame, ensuring the relative position accuracy and collaborative operation reliability of each functional module during high-speed movement and high-load grinding.

[0095] The water circulation device is integrated inside the workstation and includes components such as a water tank, water pump, pipelines, and nozzles. When the grinding operation starts, the water circulation device works synchronously, precisely spraying coolant into the grinding area through the nozzles. This effectively absorbs the frictional heat generated during the grinding process, reduces the temperature of the workpiece surface and the grinding tool 22, prevents the workpiece from undergoing thermal deformation or other changes due to local overheating, and delays the aging and wear of grinding consumables, extending the tool's service life. The coolant is collected, filtered, and recycled, which saves water resources and reduces waste liquid discharge.

[0096] The automatic consumable replacement device is an important supplement to the grinding unit 2. This device monitors the wear status of the grinding consumables in real time through sensors. When the wear of the consumables is detected to reach a preset threshold or after a certain number of grinding cycles are completed, the replacement process is automatically triggered. Its mechanical structure can accurately grasp and replace the appropriate grinding head, sandpaper and other consumables. The entire process does not require manual intervention, ensuring the continuity and consistency of the grinding operation. At the same time, the machine vision system 10 records the time, type and number of workpieces to be ground for each consumable replacement, providing data support for consumable management and process optimization.

[0097] The exhaust dust removal device is installed around the grinding area. During the grinding process, the device quickly draws the generated dust, metal shavings, and fumes into the dust removal pipe. After multi-stage filtration, clean air is discharged outdoors, while the dust is collected in a dedicated dust collection container. This effectively reduces the spread of grinding dust in the working environment, improves the working conditions of operators, reduces the risk of secondary pollution of equipment and workpieces by dust, and ensures that the workstation can operate stably for a long time in production scenarios with high requirements for environmental cleanliness.

[0098] In one embodiment, the intelligent and efficient grinding robot workstation applies an intelligent grinding control method, which corresponds one-to-one with the intelligent and efficient grinding robot workstation described in the above embodiments. For example... Figure 3 As shown, this intelligent polishing control method specifically includes the following steps:

[0099] S10: The workpiece is scanned and photographed by a machine vision system fixedly installed above the inspection area. Based on deep learning technology, the defects are reconstructed in three dimensions, and quantitative parameters of defects, including defect type, location, area, height, volume and severity, are extracted.

[0100] In this embodiment, deep learning technology refers to the ability of a model to automatically learn and recognize complex surface defect features through training on a large number of labeled defect samples based on a multi-layer neural network structure; 3D defect reconstruction refers to the process of using a deep learning model to process 2D or 3D image data, reconstruct the 3D shape of the defect area, and calculate its geometric features; and defect quantification parameters refer to a multi-dimensional data set used to comprehensively describe the surface defect features of a workpiece, including defect type classification (such as scratches, pits, pores, burrs, etc.), precise position coordinates in the workpiece coordinate system, projected area of ​​the defect area, maximum height of the defect relative to the workpiece surface, 3D spatial volume occupied by the defect, and severity level comprehensively evaluated based on multiple geometric features.

[0101] Specifically, after the workpiece enters the inspection area, compressed air is first introduced into the dust cover to maintain positive pressure and block dust interference. Then, a 3D or 2D camera fixedly installed above the inspection area scans and captures images of the workpiece surface, acquiring image data and transmitting it to the deep learning processing unit. The deep learning model identifies and segments the images, determines defect areas, and classifies defect types. Next, the defect areas are reconstructed in three dimensions, and their geometric parameters such as contour, area, height, volume, and curvature are calculated. Finally, the geometric parameters are compared with preset thresholds to assess the severity level of the defects, and a set of quantitative defect parameters containing defect type, location coordinates, geometric parameters, and severity level is integrated and output to the subsequent process matching stage.

[0102] S20: Establish a grinding process parameter library covering different materials and defect types. The process parameters include grinding angle, pressure, trajectory, number of passes, moving speed, feed rate, and dry / wet grinding modes.

[0103] In this embodiment, the grinding process parameter library refers to a database that stores the optimal grinding parameters under different working conditions. The process parameters specifically include the contact angle between the grinding tool and the workpiece surface, the applied grinding pressure, the grinding head movement trajectory, the number of grinding passes per pass, the moving speed, the feed rate, and whether to use a dry or wet grinding mode.

[0104] Specifically, the process begins by collecting historical grinding operation data and process test data, systematically organizing and calibrating the grinding process parameters for workpieces of different materials under different defect types. Through extensive experimental verification, the optimal parameter combinations for each working condition are determined, such as the best grinding angle, pressure range, and trajectory planning for pit defects in cast iron parts, and the appropriate moving speed and feed rate for scratches on aluminum alloys. These verified process parameters are then categorized and stored according to workpiece material and defect type, forming a structured process parameter library. In this library, each record corresponds to a specific "material-defect" combination, containing a complete set of process parameters, and supports dynamic updates and optimizations based on actual grinding results. This process parameter library provides the data foundation for subsequent intelligent algorithms to adaptively match the optimal process parameters.

[0105] S30: Through intelligent algorithms, the defect quantification parameters are comprehensively analyzed, and the optimal combination of process parameters is adaptively selected from the process parameter library.

[0106] In this embodiment, the intelligent algorithm refers to a matching algorithm based on similarity calculation and parameter adjustment; the optimal combination of process parameters refers to the complete set of parameters such as grinding angle, pressure, trajectory, number of passes, moving speed, feed rate and dry and wet grinding mode obtained by matching and optimizing the parameters from the process parameter library based on the current workpiece defect characteristics.

[0107] Specifically, the defect type, location, area, height, volume, and severity extracted in step S10 are first used as input feature vectors. The intelligent algorithm uses these feature vectors as a benchmark to traverse each "material-defect" record stored in the process parameter library and calculates the similarity score between the current defect feature and each case in the library. The top few records with the highest similarity are selected as candidate parameter sets. The intelligent algorithm adjusts the candidate parameters linearly or non-linearly according to the severity level of the current defect. For example, when the defect severity is high, the grinding pressure and number of passes are appropriately increased, and when the defect area is large, the trajectory coverage is adjusted. Finally, a set of optimal process parameter combinations for the current workpiece is generated and output to the grinding execution unit, realizing adaptive matching from defect identification to process decision-making.

[0108] S40: The product clamping fixture is driven by the moving slide module to switch the workpiece from the inspection area to the grinding area.

[0109] In this embodiment, the moving slide module refers to a precision displacement platform driven by a servo motor and moving along a linear guide rail. The product clamping fixture refers to a special clamping device used to fix the workpiece. The inspection area and the grinding area are two independent functional areas within the workstation, where a machine vision system and a grinding robot are deployed respectively.

[0110] Specifically, after step S10 completes workpiece inspection and generates defect quantification parameters, and step S30 completes process parameter matching, the control system sends a displacement command to the moving slide module. The moving slide module starts the servo motor, which drives the product clamping fixture mounted on the slide to move smoothly along the linear guide rail via a ball screw or synchronous belt drive. During the movement, the position sensor provides real-time feedback on the slide position to ensure that the workpiece moves accurately from the inspection area to the designated coordinate position in the grinding area. After reaching the predetermined position, the slide module automatically locks to ensure the positional stability of the workpiece during the grinding operation, providing an accurate workpiece reference for the subsequent processing operation of the grinding robot.

[0111] S50: A high-rigidity, high-load grinding robot performs grinding operations under the flexible buffer control of a force control device, based on adaptively matched process parameters and planned trajectories.

[0112] In this embodiment, the high-rigidity, high-load grinding robot refers to an industrial robot with high structural rigidity and load-bearing capacity, used to carry grinding tools and perform precision movements; the force control device refers to a force / position hybrid control unit installed at the end of the robot with flexible buffering function, used to adjust the grinding contact force in real time; the planned trajectory refers to a local fixed-point grinding path generated based on the defect location and contour.

[0113] Specifically, the control system of the grinding robot receives the optimal combination of process parameters output in step S30 and the workpiece position information determined in step S40. Based on the defect location coordinates and contour features, it plans a local fixed-point grinding trajectory to ensure that the grinding tool only covers the defect area and its neighborhood, avoiding ineffective processing of non-defect areas. It sends the matched grinding pressure value to the force control device, which uses built-in sensors to detect the contact force in real time during grinding and drives a flexible buffer mechanism to maintain a constant output force, achieving soft-contact grinding and avoiding damage to the workpiece from rigid impacts. During grinding, the force control device continuously feeds real-time contact force data back to the robot controller. The controller dynamically adjusts the robot's feed speed and attitude angle based on the deviation between the actual contact force and the target pressure, forming a closed-loop pressure control to ensure that the grinding pressure remains stable within the preset range. The robot moves along the planned trajectory, driving the grinding tool to precisely process the defect area. Throughout the grinding process, a water circulation device and a dust extraction device are simultaneously activated for cooling and dust removal. An automatic consumable replacement device replaces grinding consumables in real time according to wear conditions, ensuring continuous and stable grinding results.

[0114] S60: After grinding, the workpiece is moved back to the inspection area for re-inspection. If it is not qualified, targeted repair grinding is carried out according to the type, location and severity of the unfinished defect until it is qualified or judged as a defective product.

[0115] In this embodiment, re-inspection refers to re-inspecting the surface of the workpiece after grinding; targeted repair grinding refers to local secondary processing only on areas with residual defects; defective products refer to workpieces that still cannot meet quality standards after multiple repairs.

[0116] Specifically, after the grinding operation is completed, the control system sends a return command to the moving slide module, which drives the product clamping fixture to precisely move the workpiece from the grinding area back to the inspection area. Under the positive pressure environment maintained by the dust cover, the machine vision system restarts to scan and photograph the ground area, acquiring surface morphology data. The surface morphology data obtained from the re-inspection is compared with the preset acceptance criteria to determine whether the defects have been completely removed. If residual defects are detected, the machine vision system extracts updated defect quantification parameters, including the location coordinates, area, height, and severity level of the residual defects. Based on these residual defect parameters, the intelligent algorithm... The repair grinding process parameters are re-matched from the process parameter library, and a local repair trajectory is generated for the residual defect area. Then, the moving slide module moves the workpiece back to the grinding area, and the grinding robot performs targeted repair grinding only on the residual defect area. The above re-inspection and repair steps are repeated, and the workpiece surface is re-inspected after each repair until the entire surface is qualified or the preset maximum number of grinding times is reached. If the maximum number of grinding times is exceeded or the severity of the defect exceeds the repairable range, the workpiece is judged as a defective product, and the machine vision system automatically records the complete defect data and grinding history for subsequent quality traceability and process optimization.

[0117] In one embodiment, in step S10, the workpiece is scanned and photographed by a machine vision system fixedly installed above the detection area. The step of realizing three-dimensional reconstruction of defects based on deep learning technology specifically includes:

[0118] S11: Compressed air is introduced into the dust cover to maintain positive pressure inside the cover and prevent external grinding dust from entering and affecting the machine vision operation.

[0119] In this embodiment, positive pressure refers to a state where the air pressure inside the dust cover is higher than the external ambient air pressure. A slightly positive pressure environment is created by continuously introducing compressed air to prevent dust particles from entering the cover.

[0120] Specifically, once the workpiece enters the inspection area and is positioned, the control system activates the compressed air source, continuously supplying clean compressed air into the dust cover through the air inlet on the dust cover. The compressed air enters at a set flow rate and pressure, making the air pressure inside the cover slightly higher than the external atmospheric pressure, forming a stable micro-positive pressure difference, for example, ensuring a positive pressure of +5Pa inside the cover. This positive pressure difference effectively prevents suspended dust, metal shavings, and other tiny particles generated in the grinding area from entering the dust cover through gaps, avoiding dust adhering to the camera lens, sensor surface, or drifting in the optical path, ensuring the clarity and stability of subsequent image acquisition.

[0121] S12: The workpiece is scanned and photographed by a 3D or 2D camera fixedly installed above the detection area to obtain image data of the workpiece surface.

[0122] In this embodiment, a 3D camera or a 2D camera refers to a visual sensor used to acquire images of the workpiece surface, which is fixedly installed directly above the detection area to ensure consistency in shooting angle and distance.

[0123] Specifically, after the positive pressure environment of the dust cover is established, the control system triggers the camera to start. If a 3D camera is used, it projects an coded grating onto the workpiece surface and collects reflected images through structured light scanning or multi-view stereo vision to obtain three-dimensional point cloud data of the workpiece surface. If a 2D camera is used, it collects grayscale or color images of the workpiece surface through high-resolution imaging. The camera takes a comprehensive picture of the workpiece surface according to the preset scanning path, covering all areas to be inspected, and the acquired image data is transmitted to the deep learning processing unit in real time for subsequent analysis.

[0124] S13: Input image data into a deep learning model to identify defect areas and classify defect types.

[0125] In this embodiment, the deep learning model refers to a pre-trained convolutional neural network model used for defect detection and classification of workpiece surface images.

[0126] Specifically, after receiving image data from the camera, the deep learning processing unit first performs image preprocessing, including denoising, enhancement, and normalization, to improve image quality. Then, the preprocessed image is input into the trained deep learning model. The model extracts image features through layer-by-layer convolution and pooling operations, uses fully connected layers to classify and locate the features, identifies defective regions in the image, and determines their type.

[0127] S14: Perform three-dimensional reconstruction of the identified defect area and calculate the geometric feature parameters of the defect, including defect contour, projected area, maximum height, volume and curvature change.

[0128] In this embodiment, three-dimensional reconstruction refers to the process of reconstructing the three-dimensional shape of the defect area using two-dimensional images or multi-view data. Geometric feature parameters are quantitative indicators that describe the three-dimensional morphology of the defect.

[0129] Specifically, for point cloud data acquired by 3D cameras, the deep learning model directly segments and fits the point cloud of the defect area to reconstruct the three-dimensional shape of the defect. For images acquired by 2D cameras, multi-view imaging or photometric stereo technology is used to calculate parallax or shadows to recover the shape and reconstruct the three-dimensional structure of the defect. After obtaining the three-dimensional model of the defect, the system calculates parameters such as its contour boundary, projected area (the area covered by the defect on the workpiece surface), maximum height (the vertical distance between the deepest or highest point of the defect and the reference plane), volume (the size of the three-dimensional space occupied by the defect), and surface curvature change (the geometric steepness of the defect area).

[0130] S15: Determine the severity level of the defect by comparing the geometric feature parameters with the preset threshold.

[0131] In this embodiment, the severity level refers to the quality level comprehensively evaluated based on the defect geometric parameters, and the preset threshold refers to the grading judgment standard set for different defect types.

[0132] Specifically, the system compares each geometric feature parameter calculated in step S114 with a preset severity grading threshold. The preset threshold is set in advance according to industry standards or process requirements. For example, for pit defects, an area less than 5 mm² and a depth less than 0.5 mm is classified as slight, an area of ​​5-20 mm² or a depth of 0.5-1.5 mm is classified as moderate, and an area greater than 20 mm² or a depth greater than 1.5 mm is classified as severe. The system integrates the comparison results of various parameters and automatically determines the severity level of each defect, such as slight, moderate, or severe, according to the preset rating rules.

[0133] S16: Combine the defect type, location coordinates, geometric feature parameters, and severity level to form a complete set of defect quantification parameters.

[0134] In this embodiment, the defect quantification parameter set refers to a multi-dimensional data set used to comprehensively describe the surface defect characteristics of the workpiece, serving as the input basis for subsequent grinding process decisions.

[0135] Specifically, the defect type identified in step S113, the geometric feature parameters calculated in step S114 (including contour, area, height, volume, and curvature), the severity level assessed in step S115, and the precise location coordinates of the defect in the workpiece coordinate system are integrated. All data is formatted and stored according to a preset data structure to form a complete defect record. If there are multiple defects on the workpiece surface, multiple records are generated to form a defect quantification parameter set. This parameter set is finally output to the subsequent process matching module for adaptive selection of the optimal grinding process parameters.

[0136] In one embodiment, step S30, which involves comprehensively analyzing the defect quantification parameters using an intelligent algorithm and adaptively selecting the optimal combination of process parameters from the process parameter library, includes:

[0137] S31: Construct a process parameter library that covers the correspondence between different workpiece materials, different defect types and grinding process parameters. The process parameters include grinding angle, pressure, trajectory, number of passes, moving speed, feed rate and dry and wet grinding modes.

[0138] In this embodiment, the process parameter library refers to a database that stores the optimal grinding parameters under different working conditions; the process parameters refer to the contact angle between the grinding tool and the workpiece surface, the applied grinding pressure, the grinding head movement trajectory, the number of grinding passes, the moving speed, the feed rate, and whether to use a dry or wet grinding mode.

[0139] Specifically, the process parameter library is built based on a large amount of experimental data and actual production cases, covering a variety of common workpiece materials (such as aluminum alloys, carbon steel, stainless steel, etc.) and typical defect types (such as scratches, oxide scale, burrs, pits, etc.). For each "material-defect" combination, the grinding process parameters are optimized through orthogonal experiments, and the grinding effects under different parameter combinations (such as surface roughness, defect removal rate, workpiece deformation, etc.) are recorded. The optimal parameter set is then selected and stored in the database. The database adopts a structured storage method, using material type and defect characteristics as the main index, which facilitates fast retrieval and matching. At the same time, the database reserves a dynamic update interface to continuously expand the parameter set based on new materials, new defects, or process optimization results in actual production, ensuring the timeliness and applicability of the parameter library.

[0140] S32: Use the extracted defect type, location, area, height, volume, and severity as input features.

[0141] In this embodiment, input features refer to multi-dimensional quantitative data used to describe the defects of the current workpiece, which serves as the basis for the intelligent algorithm to match process parameters.

[0142] Specifically, the defect information of the current workpiece is read from the defect quantification parameter set output in step S10, including the defect type (such as scratches, dents, etc.), the precise position coordinates in the workpiece coordinate system, the projected area of ​​the defect region, the maximum height of the defect relative to the workpiece surface, the three-dimensional volume occupied by the defect, and the comprehensive severity level (such as minor, moderate, severe). These parameters are organized according to the preset data format to form an input feature vector for subsequent intelligent algorithm matching and analysis.

[0143] S33: The input features are comprehensively analyzed through intelligent algorithms to calculate the matching degree between the current defect features and each case in the process parameter library.

[0144] In this embodiment, the intelligent algorithm refers to a matching algorithm based on similarity calculation, where the matching degree refers to the degree of similarity between the current defect feature and each case in the process parameter library.

[0145] Specifically, the intelligent algorithm uses the input feature vector formed in step S312 as a benchmark and traverses each "material-defect" record stored in the process parameter library. The algorithm first normalizes the input features and the features of each case in the library to eliminate the influence of different dimensions. Then, it uses Euclidean distance or cosine similarity calculation method to calculate the similarity score between the current defect feature and the feature of each case. For a workpiece containing multiple defects, the algorithm calculates the matching degree between each defect and each case, and performs weighted summation according to the severity of the defects to obtain the overall matching degree.

[0146] S34: Match the process parameter combination with the highest similarity from the process parameter library as the basic parameter.

[0147] In this embodiment, the basic parameters refer to the combination of process parameters that is most similar to the current defect characteristics and is matched from the process parameter library.

[0148] Specifically, the intelligent algorithm sorts all the matching scores calculated in step S313 and selects the top few records with the highest similarity as a candidate parameter set; from the candidate parameter set, it selects the record with the highest matching degree and extracts its corresponding complete process parameter combination, including grinding angle, pressure, trajectory, number of passes, moving speed, feed rate, and dry / wet grinding mode, as the basic parameters for subsequent adjustments; if there are multiple defects and the optimal parameters for each defect are different, the system comprehensively weighs the defect location distribution and severity to determine a unified basic parameter or to use different parameters for different regions.

[0149] S35: Adjust the basic parameters linearly or nonlinearly according to the severity of the defect to generate the optimal combination of process parameters.

[0150] In this embodiment, the optimal combination of process parameters refers to the most suitable process parameters obtained by dynamically adjusting the basic parameters based on the current defect characteristics of the workpiece.

[0151] Specifically, the intelligent algorithm optimizes and adjusts the basic parameters obtained in step S314 based on the severity level of the current defect. For defects with high severity, a non-linear adjustment strategy is adopted, such as increasing the grinding pressure by 30%-50%, increasing the number of grinding passes by 1-2, reducing the moving speed by 20%-30%, and appropriately reducing the feed rate to ensure complete removal of the defect. For defects with low severity, a linear adjustment strategy is adopted, such as slightly increasing the pressure, keeping the number of passes unchanged, or making minor adjustments to avoid over-grinding. The algorithm adjusts the trajectory coverage and density based on the defect area and volume, and optimizes the trajectory starting point and path based on the defect location coordinates. Finally, a set of optimal process parameters for the current workpiece is generated, including precise grinding angle, pressure value, planned trajectory, number of grinding passes, moving speed, feed rate, and dry / wet mode, and outputs it to the grinding unit.

[0152] In one embodiment, step S50, where the high-rigidity, high-load grinding robot performs the grinding operation under the flexible buffer control of the force control device according to adaptively matched process parameters and planned trajectory, includes:

[0153] S51: Based on the location coordinates and outline of the defect, plan a local fixed-point grinding trajectory for the defect area, so that the grinding tool only covers the defect area and its neighborhood, avoiding full grinding of non-defect areas.

[0154] In this embodiment, the local fixed-point grinding trajectory refers to a grinding path that only covers the defect area and its surrounding small neighborhood. The neighborhood refers to the area that extends outward from the defect boundary by a certain width, which is used to ensure that the defect edge is completely removed.

[0155] Specifically, the control system receives the defect location coordinates and contour data output in step S10, as well as the process parameters matched in step S30; it generates a basic grinding area based on the defect contour, and then expands it outward by a preset neighborhood width (e.g., 2-5mm) to form the final grinding coverage area; it plans a specific grinding trajectory based on the defect shape and size, using a spiral trajectory for circular or elliptical defects and a reciprocating straight trajectory for elongated defects to ensure that the grinding tool evenly covers the entire defect area; after the trajectory is generated, the system performs collision detection and inspection to ensure that the trajectory is safe and feasible.

[0156] S52: Sends the matching grinding pressure value to the force control device, so that the force control device maintains a constant output force during the grinding process, realizes flexible buffer control, and avoids hard collisions that damage the workpiece.

[0157] In this embodiment, constant output force refers to the target grinding pressure value set according to process requirements.

[0158] Specifically, the control system sends the grinding pressure value matched in step S30 to the controller of the force control device via the communication bus. The force control device has a built-in high-precision force sensor and servo drive mechanism. Before the grinding tool contacts the workpiece surface, the force control mode is switched to constant force control. When the grinding tool contacts the workpiece, the force control device automatically adjusts the displacement through the built-in flexible buffer mechanism (such as spring, cylinder or electromagnetic drive) according to the deviation between the real-time detected contact force and the target pressure, so that the actual contact force is always maintained near the target pressure, realizing flexible contact and avoiding rigid impact caused by robot positioning error or uneven workpiece surface.

[0159] S53: The force control device provides real-time feedback of contact force data. The high-rigidity, high-load grinding robot dynamically adjusts its feed speed and attitude angle based on the deviation between the real-time contact force and the target pressure, thus precisely controlling the grinding pressure.

[0160] In this embodiment, dynamic adjustment refers to the process by which the robot performs closed-loop adjustment of motion parameters based on real-time data fed back by the force control device.

[0161] Specifically, the force control device collects actual contact force data at millisecond-level frequency and feeds it back to the robot controller via a real-time communication bus. The robot controller compares the actual contact force with the target pressure and calculates the deviation. When the actual contact force is higher than the target pressure, the controller appropriately reduces the robot's feed speed or adjusts the attitude angle to reduce the normal component, causing the contact force to fall back. When the actual contact force is lower than the target pressure, the controller appropriately increases the feed speed or adjusts the attitude angle to increase the normal component, thereby increasing the contact force. This ensures that the actual contact force remains stable within the allowable error range of the target pressure throughout the entire grinding process, achieving high-precision constant force grinding.

[0162] S54: Simultaneously start the water circulation device and the dust extraction device for cooling and dust removal.

[0163] S55: The automatic consumable replacement device replaces the polishing consumables in real time according to the wear condition of the consumables or the preset replacement cycle to ensure the polishing effect.

[0164] In this embodiment, consumable wear refers to the remaining service life or degree of wear of the grinding tool.

[0165] Specifically, when the robot begins grinding, the control system simultaneously sends start commands to the water circulation device and the dust extraction device. The water circulation device starts the water pump and precisely sprays coolant into the grinding area through nozzles to absorb the frictional heat generated during grinding, reducing the temperature of the workpiece surface and the grinding tool. The coolant is collected, filtered, and recycled. The dust extraction device starts the fan and creates negative pressure in the grinding area through the dust suction hood, quickly drawing the generated dust into the air duct. After being filtered by a high-efficiency filter, the dust is discharged and collected in a dust collection container. The automatic consumable replacement device has a built-in wear detection sensor that monitors the wear status of the grinding tool in real time during the grinding process, such as the remaining abrasive thickness and changes in grinding efficiency. When wear is detected to reach a preset threshold, or according to a preset replacement cycle (e.g., every 10 workpieces), the control system issues a replacement command. The device automatically grabs a new grinding tool from the consumable storage bin using a mechanical gripper, quickly completes the replacement operation, and records the replacement information in the system log. During the replacement process, the robot pauses its operation and moves to the designated tool changing position. After the replacement is completed, grinding automatically resumes, minimizing downtime and ensuring consistent grinding results.

[0166] In one embodiment, step S60, i.e., after grinding is completed, the workpiece is moved back to the inspection area for re-inspection; if it fails, targeted repair grinding is performed.

[0167] S61: After grinding, the workpiece is moved from the grinding area back to the inspection area by moving the slide module;

[0168] S62: Under the condition of maintaining positive pressure of the dust cover, the polished area is scanned and inspected again by the machine vision system fixedly installed above the inspection area to obtain the surface morphology data after polishing.

[0169] In this embodiment, surface morphology data refers to the three-dimensional geometric information of the workpiece surface after grinding, which is used to evaluate the grinding quality.

[0170] Specifically, after completing the grinding operation, the moving slide module moves precisely along the preset return trajectory, smoothly sending the workpiece back to the designated position in the inspection area. The dust cover continuously maintains a positive pressure environment, effectively preventing external dust from entering the inspection area and ensuring the cleanliness of the inspection environment. The machine vision system restarts, using a high-resolution camera and structured light projection device to perform a comprehensive scan of the ground area, acquiring three-dimensional geometric morphology data, including surface roughness, flatness, and contour accuracy. The system compares and analyzes the collected morphology data with preset quality standards to quickly assess whether the grinding effect meets the standards.

[0171] S63: Compare the surface morphology data obtained from the re-inspection with the preset acceptance standard to determine whether the defects have been completely removed;

[0172] S64: If a residual defect is detected, extract the quantitative parameters of the residual defect, including the updated location, area, height, and severity.

[0173] In this embodiment, the preset acceptance standard refers to the quantitative indicators for the acceptance of workpiece surface quality, including parameters such as the maximum allowable residual defect size and surface roughness; residual defects refer to defects that are not completely removed after one grinding, and their quantitative parameters are used to guide restorative grinding.

[0174] Specifically, the control system compares the surface morphology data obtained from the re-inspection with the preset acceptance standards item by item, such as whether the surface roughness is lower than Ra0.8μm and whether the maximum height of the defect area is less than 0.05mm. If all indicators meet the requirements, the grinding is deemed qualified and proceeds to the next process. If any indicator fails to meet the standard, the system automatically locates the residual defect area through intelligent algorithms and extracts its quantitative parameters, including the precise position coordinates of the defect in the workpiece coordinate system, the projected area, the maximum height, the three-dimensional volume, and the severity level reassessed according to preset rules (such as slight, moderate, severe). These parameters will serve as the input basis for restorative grinding to ensure that the secondary grinding is accurately targeted at the substandard areas.

[0175] S65: Based on the residual defect quantification parameters, the repair grinding process parameters are re-matched from the process parameter library through intelligent algorithms;

[0176] S66: Move the workpiece back to the grinding area and perform targeted repair grinding only on the remaining defective areas;

[0177] S67: Repeat the re-inspection and repair steps until the workpiece is qualified or the preset maximum number of grinding times is reached;

[0178] S68: If the maximum number of polishing cycles is exceeded or the severity of the defect exceeds the repairable range, the product is judged as defective and the defect data and polishing history are recorded.

[0179] In this embodiment, the restorative grinding process parameters refer to the secondary grinding parameters optimized for the characteristics of residual defects; the fixed-point restorative grinding refers to the secondary processing of only the residual defect area to avoid repeated damage to the qualified area; the maximum number of grinding times refers to the maximum number of grinding times allowed for a single workpiece to prevent over-grinding; the defective product refers to the workpiece that still cannot meet the quality standard after multiple repairs; and the repairable range refers to the upper limit of the defects that can be repaired according to the process capability.

[0180] Specifically, based on the residual defect quantification parameters extracted in step S614, the control system again invokes the intelligent algorithm to select the most suitable combination of process parameters for repairing the current residual defect from the process parameter library. Similar to the initial grinding parameter matching, the intelligent algorithm normalizes the features of the residual defect and calculates its similarity score with each case in the library, selecting the parameter with the highest matching degree as the basic parameter for repair grinding. Considering that residual defects are usually small in area and low in severity, the intelligent algorithm finely adjusts the basic parameters, such as appropriately reducing grinding pressure, reducing the number of grinding passes, and narrowing the trajectory coverage area to avoid over-grinding and damaging the workpiece surface. The adjusted repair grinding process parameters are sent to the grinding robot, and at the same time, based on the precise location coordinates of the residual defect, the local fixed-point grinding trajectory is replanned to ensure that the grinding tool only covers the residual defect area and its surrounding area. In a small neighborhood, the moving slide module moves the workpiece from the inspection area back to the grinding area. Under the flexible buffer control of the force control device, the high-rigidity, high-load grinding robot performs targeted repair grinding on the residual defect area according to the new process parameters and trajectory. After grinding, the re-inspection process of steps S611 to S614 is repeated to check the surface morphology data of the grinding area again to determine whether the defects have been completely removed. If there are still residual defects, the repair grinding steps continue until the surface quality of the workpiece meets the preset qualified standard or reaches the preset maximum number of grinding times. If the maximum number of grinding times is exceeded or the severity of the defects exceeds the repairable range, the control system automatically determines that the workpiece is a defective product and records the defect data (such as defect type, location, area, height, etc.) and grinding history (such as number of grinding times, process parameters used, etc.) in detail, providing data support for subsequent process optimization and quality traceability.

[0181] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0182] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention. The actual structure is not limited to this. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. An intelligent and efficient grinding robot workstation, characterized in that, include: The detection unit (1) includes a machine vision system (10) and a dust cover (11); the machine vision system (10) uses a 3D camera or a 2D camera, which is fixedly installed above the detection area to detect the defect features of the workpiece; the dust cover (11) is filled with positive pressure gas to prevent external dust from entering and affecting the operation of the machine vision system (10) during the detection process. The grinding unit (2) includes a high-rigidity, high-load grinding robot (20), a force control device (21), grinding tools (22), and consumables. The high-rigidity, high-load grinding robot (20) is used to carry the grinding tools (22) and execute the grinding trajectory. The force control device (21) has a flexible buffer function and is installed at the end of the grinding robot (20) to precisely control the grinding pressure. The grinding tools (22) and consumables are selected and adapted according to the workpiece material and defect type. The moving unit (3) includes a product clamping fixture (30) and a moving slide module (31); the product clamping fixture (30) is used to fix the workpiece; the moving slide module (31) is connected to the product clamping fixture (30) and is used to drive the product clamping fixture (30) to switch between the inspection area and the grinding area. The control system is communicatively connected to the detection unit (1), the grinding unit (2) and the moving unit (3), and is used to control the grinding actuator to perform grinding operations on the workpiece based on the quantitative parameters of the defect.

2. The intelligent and efficient grinding robot workstation according to claim 1, characterized in that, The machine vision system (10) also includes a deep learning processing unit, which realizes three-dimensional reconstruction of defects based on deep learning technology and extracts defect quantification parameters including defect type, location, area, height, volume and severity.

3. The intelligent and efficient grinding robot workstation according to claim 1, characterized in that, The dust cover (11) is provided with a compressed air inlet for introducing compressed air into the dust cover (11) and maintaining positive pressure inside the cover.

4. The intelligent and efficient grinding robot workstation according to claim 1, characterized in that, It also includes auxiliary functional units, which include a frame, a water circulation device, an automatic consumable replacement device, and a dust extraction device. The rack is used to support and install the various functional modules; The water circulation device is used for cooling during the grinding process; The automatic consumable replacement device is used to replace the polishing consumables in real time according to the wear condition of the consumables. The exhaust dust removal device is used to handle the dust generated during grinding.

5. The intelligent and efficient grinding robot workstation according to claim 1, wherein the intelligent and efficient grinding robot workstation applies an intelligent grinding control method, the intelligent grinding control method comprising the steps of: The workpiece is scanned and photographed by a machine vision system fixedly installed above the inspection area. Based on deep learning technology, the defects are reconstructed in three dimensions, and quantitative parameters of defects, including defect type, location, area, height, volume and severity, are extracted. Establish a grinding process parameter library covering different materials and defect types. The process parameters include grinding angle, pressure, trajectory, number of passes, moving speed, feed rate, and dry / wet grinding modes. By comprehensively analyzing the defect quantification parameters through intelligent algorithms, the optimal combination of process parameters is adaptively selected from the process parameter library; The workpiece is switched from the inspection area to the grinding area by moving the product clamping fixture through the sliding table module; The high-rigidity, high-load grinding robot performs grinding operations under the flexible buffer control of the force control device, based on adaptively matched process parameters and planned trajectories. After grinding, the workpiece is moved back to the inspection area for re-inspection. If it fails, targeted repair grinding is carried out according to the type, location and severity of the unfinished defects until it passes or is judged as a defective product.

6. The intelligent and efficient grinding robot workstation according to claim 5, characterized in that, The steps of scanning and photographing the workpiece using a machine vision system fixedly installed above the inspection area, and reconstructing the three-dimensional defects based on deep learning technology, specifically include: Compressed air is introduced into the dust cover to maintain positive pressure inside the cover and prevent external grinding dust from entering and affecting the machine vision operation. The workpiece is scanned and photographed by a 3D or 2D camera that is fixedly installed above the inspection area to obtain image data of the workpiece surface. Image data is input into a deep learning model to identify defect areas and classify defect types. The identified defect areas are reconstructed in three dimensions, and the geometric feature parameters of the defects are calculated, including the defect contour, projected area, maximum height, volume and curvature change. The severity level of the defect is determined by comparing the geometric feature parameters with a preset threshold. The defect type, location coordinates, geometric feature parameters, and severity level are combined to form a complete set of defect quantification parameters.

7. The intelligent and efficient grinding robot workstation according to claim 5, characterized in that, The step of comprehensively analyzing defect quantification parameters using intelligent algorithms and adaptively selecting the optimal combination of process parameters from the process parameter library includes: Construct a process parameter library that covers the correspondence between different workpiece materials, different defect types and grinding process parameters. The process parameters include grinding angle, pressure, trajectory, number of passes, moving speed, feed rate and dry and wet grinding modes. The extracted defect type, location, area, height, volume, and severity are used as input features; The input features are comprehensively analyzed by intelligent algorithms to calculate the matching degree between the current defect features and each case in the process parameter library; The process parameter combination with the highest similarity is selected from the process parameter library as the basic parameter; The basic parameters are adjusted linearly or nonlinearly based on the severity of the defects to generate the optimal combination of process parameters.

8. The intelligent and efficient grinding robot workstation according to claim 5, characterized in that, The high-rigidity, high-load grinding robot performs the grinding operation according to adaptively matched process parameters and planned trajectory, under the flexible buffer control of the force control device. The steps include: Based on the location coordinates and outline of the defect, a local fixed-point grinding trajectory is planned for the defect area, so that the grinding tool only covers the defect area and its neighborhood, avoiding full grinding of non-defect areas. The matching grinding pressure value is sent to the force control device, so that the force control device maintains a constant output force during the grinding process, realizing flexible buffer control and avoiding hard collisions that damage the workpiece. The force control device provides real-time feedback of contact force data. The high-rigidity, high-load grinding robot dynamically adjusts its feed speed and attitude angle based on the deviation between the real-time contact force and the target pressure, thus precisely controlling the grinding pressure. Simultaneously start the water circulation device and the dust extraction and removal device for cooling and dust treatment; The automatic consumable replacement device replaces the polishing consumables in real time according to the wear condition of the consumables or the preset replacement cycle to ensure the polishing effect.

9. The intelligent and efficient grinding robot workstation according to claim 5, characterized in that, After grinding, the workpiece is moved back to the inspection area for re-inspection. If it fails, targeted repair grinding is performed. After grinding, the workpiece is moved from the grinding area back to the inspection area using the movable slide module; Under the condition of maintaining positive pressure of the dust cover, the polished area is scanned and inspected again by the machine vision system fixedly installed above the inspection area to obtain the surface morphology data after polishing. The surface morphology data obtained from the re-inspection is compared with the preset acceptance standard to determine whether the defects have been completely removed; If a residual defect is detected, extract the quantitative parameters of the residual defect, including the updated location, area, height, and severity. Based on the residual defect quantification parameters, the repair grinding process parameters are re-matched from the process parameter library using an intelligent algorithm; The workpiece is moved back to the grinding area, and only the areas with residual defects are ground for targeted repair grinding. Repeat the re-inspection and repair steps until the workpiece is qualified or the preset maximum number of grinding times is reached; If the maximum number of polishing cycles is exceeded or the severity of the defect exceeds the repairable range, the product is judged as defective and the defect data and polishing history are recorded.

Images