A curved surface part polishing and grinding device

Abstract

This invention discloses a polishing and grinding equipment for curved surfaces, relating to the field of precision surface treatment technology. The equipment includes modules for curved surface data acquisition, digital twin modeling, surface energy field analysis, process parameter generation, multiphysics polishing, quality prediction, real-time feedback, a self-evolving process library, and cloud-edge collaborative decision-making. By acquiring the three-dimensional morphology, material, and defect information of the curved surface part, a digital twin model is constructed, and differentiated polishing strategies are generated based on the surface energy field distribution. High-precision surface treatment is achieved through the synergistic effect of multiphysics fields. Simultaneously, a closed-loop control mechanism is formed through quality prediction and real-time feedback, and the processing parameters are continuously optimized using a self-evolving process library, achieving autonomous learning and iterative optimization of the process. This equipment can significantly improve the processing consistency, surface quality, and process adaptability of complex curved surface parts.

Description

Technical Field

[0001] This application relates to the field of precision surface treatment technology, and in particular to a polishing and grinding equipment for curved parts. Background Technology

[0002] Curved surfaces are widely used in aerospace, precision molds, medical devices, and high-end equipment manufacturing. Their surface quality directly affects their performance and service life. Current polishing processes for curved surfaces typically rely on preset process parameters, making it difficult to adjust polishing strategies in real-time based on changes in local curvature, surface defect distribution, and processing conditions. This leads to insufficient processing consistency in complex curved surface areas, consequently affecting polishing accuracy and surface quality stability. Therefore, achieving dynamic optimization and consistent quality control in the polishing process of curved surfaces has become a pressing technical problem to be solved in this field.

[0003] Therefore, a polishing and grinding device for curved parts is invented to solve the problems mentioned in the background art. Summary of the Invention

[0004] To address the problem in existing technologies where it is difficult to adjust process parameters in real time according to the processing status during the polishing process of curved parts, resulting in insufficient processing consistency in complex curved areas, this invention provides a polishing and grinding device for curved parts.

[0005] This application provides a polishing and grinding equipment for curved surface parts, which adopts the following technical solution: include: Surface data acquisition module; Digital twin modeling module; Surface energy field analysis module; Process parameter generation module; Multiphysics polishing module; Quality prediction module; Self-evolving process library module; Cloud-edge collaborative decision-making module; Real-time feedback module; in, The surface data acquisition module is used to acquire the three-dimensional contour data, surface roughness data, material parameter data, and defect distribution data of the surface part to be processed; The digital twin modeling module is used to construct a digital twin model of the part to be processed based on the acquired data; The surface energy field analysis module is used to establish a curved surface energy distribution model and generate a surface energy field distribution map. The process parameter generation module generates differentiated polishing parameters for the corresponding region based on the digital twin model and the surface energy field distribution results; The multi-physics polishing module is used to complete the surface polishing process by utilizing the synergistic effect of at least two physical fields among electric field, magnetic field, ultrasonic field and plasma field; The quality prediction module is used to predict the surface roughness, material removal amount, and residual stress state after polishing in real time. The real-time feedback module is used to collect actual processing results and compare them with predicted results; The self-evolving process library module updates the process feature data based on the comparison results; The cloud-edge collaborative decision-making module calls the updated process feature data to generate the next round of optimized processing strategy.

[0006] Optionally, the digital twin modeling module includes a geometry mapping unit, a material mapping unit, a defect mapping unit, and a processing state mapping unit.

[0007] Optionally, the surface energy field analysis module uses curvature distribution parameters, surface free energy parameters, and defect density parameters to construct a curved surface energy field model.

[0008] Optionally, the process parameter generation module uses a reinforcement learning algorithm to establish a mapping relationship between processing parameters and surface quality.

[0009] Optionally, the multiphysics polishing module includes at least two of the following: a magnetorheological processing unit, an ultrasonic strengthening processing unit, an electrochemical processing unit, and a low-temperature plasma processing unit.

[0010] Optionally, the quality prediction module uses a neural network prediction model to predict the material removal rate in real time.

[0011] Optionally, the real-time feedback module includes a surface profile detection unit, a roughness detection unit, and a stress detection unit.

[0012] Optionally, the self-evolving process library module uses a case-based reasoning algorithm and an incremental learning algorithm to update process data.

[0013] Optionally, the cloud-edge collaborative decision-making module uses edge computing nodes for real-time decision-making and optimizes parameters by training models in the cloud.

[0014] Optionally, the device may also include a process knowledge graph module for establishing the relationship between part material, curvature characteristics, defect type and process parameters.

[0015] In summary, this application includes the following beneficial technical effects: 1. This invention constructs a virtual mapping model of curved surface parts through a digital twin modeling module, which can predict and analyze the surface features and processing status before processing, thereby improving the accuracy of process parameter matching.

[0016] 2. This invention uses a surface energy field analysis module to evaluate the energy characteristics of different regions of a curved surface, enabling differentiated process parameter configuration and improving the processing uniformity of complex curved surface regions.

[0017] 3. The present invention uses a multi-physics field synergistic polishing method for surface treatment, which can reduce the risk of local over-processing caused by traditional contact processing and improve the stability of surface quality.

[0018] 4. This invention constructs a closed-loop control system for processing through a quality prediction module and a real-time feedback module, thereby achieving dynamic correction and real-time optimization during the processing and improving polishing accuracy.

[0019] 5. This invention features a self-evolving process library module, which continuously learns from historical processing data and results to achieve autonomous optimization of process parameters, thereby improving the equipment's adaptability to parts with different curved surfaces.

[0020] 6. This invention forms an intelligent processing link of "digital twin modeling - surface energy field analysis - quality prediction - real-time feedback - process self-evolution", which can effectively improve the quality consistency, processing efficiency and intelligence level of the polishing process of complex curved parts. Attached Figure Description

[0021] Figure 1 This is a block diagram showing the overall modular connection of the equipment in this device; Figure 2 This is a schematic diagram of the curved surface data acquisition module of this device; Figure 3 This is a schematic diagram of the digital twin modeling module structure of this device; Figure 4 This is a schematic diagram of the surface energy field analysis module of this device; Figure 5 This is a schematic diagram of the multiphysics polishing module structure of this device; Figure 6 This is a flowchart of the polishing and grinding method for curved parts in this device; 1. Surface data acquisition module; 2. Digital twin modeling module; 3. Surface energy field analysis module; 4. Process parameter generation module; 5. Multiphysics polishing module; 6. Quality prediction module; 7. Real-time feedback module; 8. Self-evolving process library module; 9. Cloud-edge collaborative decision-making module. Detailed Implementation

[0022] The present invention will be further described below with reference to specific implementation examples and accompanying drawings, but the present invention is not limited to these embodiments. Example

[0023] like Figures 1 to 6As shown, this invention discloses a polishing method for a polishing and grinding equipment for curved surfaces, the method comprising: Step S1: Collect the global feature data of the surface part to be processed through the surface data acquisition module 1; the global feature data includes three-dimensional contour data, surface roughness data, material parameter data and defect distribution data.

[0024] Step S2: Construct a full-domain digital twin model of the part based on full-domain feature data through the digital twin modeling module 2, and complete the four-dimensional mapping of geometry, material, defects, and processing status.

[0025] Step S3: Extract the core feature parameters of the curved surface through the surface energy field analysis module 3, construct a quantitative model of the surface energy field of the curved surface, and generate a global surface energy field distribution map.

[0026] Step S4: Using the reinforcement learning algorithm, the process parameter generation module 4 generates differentiated polishing process parameters for curved sub-regions by combining the digital twin model and the surface energy field distribution results.

[0027] Step S5: The multi-physics polishing module 5 employs at least two physical fields in synergy to perform differential polishing and grinding operations on curved surfaces.

[0028] Step S6: Using the quality prediction module 6 and a neural network prediction model, the surface roughness, material removal amount, and residual stress state after polishing are predicted in real time.

[0029] Step S7: Collect actual processing quality data through real-time feedback module 7, compare the deviation with the predicted data, and generate a processing error dataset.

[0030] Step S8: The process feature database is iteratively updated based on the error dataset by combining the case reasoning algorithm and the incremental learning algorithm with the self-evolving process library module 8.

[0031] Step S9: The updated process data is called through the cloud-edge collaborative decision-making module 9 to generate the next round of optimized processing strategy, thereby realizing closed-loop iterative polishing processing.

[0032] The following is a detailed discussion of each step: Step S1: Collect the global feature data of the surface part to be processed through the surface data acquisition module 1; the global feature data includes three-dimensional contour data, surface roughness data, material parameter data and defect distribution data.

[0033] Step S1 specifically includes: The surface data acquisition module 1 integrates a three-dimensional contour detection unit, a roughness detection unit, a material detection unit, and a defect detection unit. It adopts a non-contact full-domain scanning method to collect data on curved parts without blind spots, and is suitable for complex curved surface structures with curvature radii of 5mm to 500mm.

[0034] The three-dimensional contour data includes the coordinates of each point on the surface, the local radius of curvature, and the surface slope parameters; the material parameter data includes the part's hardness, elastic modulus, thermal conductivity, and material density; the defect distribution data includes the defect location coordinates, defect type, defect area, and defect density.

[0035] Define the local curvature parameter K of the surface as follows:

[0036] In the formula, y' is the first derivative of the surface curve, y'' is the second derivative of the surface curve, and K is the local curvature of the surface. The larger the curvature value, the higher the degree of bending of the surface and the greater the difficulty of polishing.

[0037] The surface defect density parameter Dd is defined as follows:

[0038] In the formula, Sdefect is the total area of ​​defects in a single region, and Stotal is the total area of ​​the curved sub-region. The larger the value of Dd, the more concentrated the defects in the region are, and the more intense the polishing removal needs to be.

[0039] In this embodiment, the data acquisition accuracy is set to 0.01 mm and the roughness acquisition accuracy is 0.001 μm to ensure that the original data can accurately reflect the true state of the surface and provide reliable data support for subsequent modeling and analysis.

[0040] Step S2: Construct a full-domain digital twin model of the part based on full-domain feature data through the digital twin modeling module 2, and complete the four-dimensional mapping of geometry, material, defects, and processing status.

[0041] Step S2 specifically includes: The digital twin modeling module 2 includes a geometric mapping unit, a material mapping unit, a defect mapping unit, and a processing state mapping unit. These four types of units work together to achieve a 1:1 accurate mapping between the virtual model and the physical part.

[0042] The geometric mapping unit calls the three-dimensional contour data and reconstructs the three-dimensional virtual model of the part through the surface fitting algorithm, accurately replicating the local curvature and contour shape of the surface, with a geometric mapping error ≤0.02mm.

[0043] The material mapping unit binds the measured material parameters to the virtual model, enabling differentiated assignment of material properties for different curved surface regions and distinguishing the processing characteristics of different materials such as cemented carbide, stainless steel, and titanium alloy.

[0044] The defect mapping unit marks the location, density, and type of defects to the corresponding points in the virtual model, forming a visual defect distribution cloud map, and accurately locating high-defect processing areas.

[0045] The processing status mapping unit reserves a real-time data interface to synchronize the equipment operating parameters, physical field intensity, and processing time of subsequent polishing processes, thereby achieving virtual-real synchronization of the processing process.

[0046] Step S3: Extract the core feature parameters of the curved surface through the surface energy field analysis module 3, construct a quantitative model of the surface energy field of the curved surface, and generate a global surface energy field distribution map.

[0047] Step S3 specifically includes: Surface energy field analysis module 3 uses the curvature distribution parameter K, surface free energy parameter Es, and defect density parameter Dd as core inputs to construct a quantitative model of the surface energy field of a curved surface. The formula for calculating the surface energy value E of a single region is as follows:

[0048] In the formula, ω1, ω2, and ω3 are weighting coefficients, and ω1+ω2+ω3=1, which are adaptively adjusted according to the material of the part; E is the surface energy value of the curved sub-region. The higher the energy value, the greater the polishing difficulty of the region and the higher the processing energy required.

[0049] The global surface is divided into several square sub-regions of equal size, with the preferred size of a single region being 1mm×1mm. The surface energy value is calculated for each region, and the regions are divided into three levels: high-energy region, medium-energy region, and low-energy region according to the energy value. Finally, a visualized global surface energy field distribution map is generated to accurately distinguish the differentiated processing areas.

[0050] Step S4: Using the reinforcement learning algorithm, the process parameter generation module 4 generates differentiated polishing process parameters for curved sub-regions by combining the digital twin model and the surface energy field distribution results.

[0051] Step S4 specifically includes: The process parameter generation module 4 incorporates a deep reinforcement learning (DQN) algorithm. It takes surface energy, material parameters, and curvature features as inputs, and aims to optimize uniform surface roughness, consistent material removal, and minimum residual stress. It establishes a dynamic mapping relationship between surface features and polishing process parameters.

[0052] The algorithm input parameters include: regional surface energy value E, local curvature K, part hardness H, and defect density Dd; the algorithm output process parameters include: polishing travel speed Vp, physical field strength Ip, processing time Tp, and abrasive particle size Dp.

[0053] The reinforcement learning reward function R is defined as:

[0054] In the formula, ΔRa is the deviation between the actual roughness and the target roughness, ΔM is the deviation of material removal amount, σ is the residual stress value, and λ1, λ2, and λ3 are weighting coefficients. The larger the reward function R is, the higher the matching accuracy of the process parameters.

[0055] For high-energy defect areas, the polishing travel speed is automatically reduced, the physical field intensity is increased, and the processing time is extended; for low-energy flat areas, the processing speed is increased and the processing energy is reduced, achieving differentiated and precise polishing across the entire area.

[0056] Step S5: The multi-physics polishing module 5 employs at least two physical fields in synergy to perform differential polishing and grinding operations on curved surfaces.

[0057] Step S5 specifically includes: The multi-physics polishing module 5 integrates a magnetorheological processing unit, an ultrasonic strengthening processing unit, an electrochemical processing unit, and a low-temperature plasma processing unit. In this embodiment, the preferred processing mode is a dual-physics field synergistic processing mode of magnetorheological field + ultrasonic field.

[0058] The magnetorheological processing unit outputs flexible polishing force, which can adapt to curved and irregular contours and avoid overcutting damage caused by rigid contact; the ultrasonic strengthening processing unit outputs high-frequency vibration energy to enhance the removal effect of trace materials. The two work synchronously and complement each other.

[0059] Physical field intensity matching rules: high energy region magnetorheological field intensity 1.2~1.5T, ultrasonic field frequency 25~30kHz; medium energy region magnetorheological field intensity 0.8~1.2T, ultrasonic field frequency 20~25kHz; low energy region magnetorheological field intensity 0.5~0.8T, ultrasonic field frequency 15~20kHz.

[0060] Step S6: Using the quality prediction module 6 and a neural network prediction model, the surface roughness, material removal amount, and residual stress state after polishing are predicted in real time.

[0061] Step S6 specifically includes: The quality prediction module 6 adopts a three-layer BP neural network prediction model. The input layer consists of real-time process parameters and surface feature parameters, the hidden layer has 24 neurons, and the output layer consists of three quality indicators: surface roughness Ra, material removal amount M, and residual stress σ.

[0062] The neural network activation function is ReLU, the loss function is mean squared error (MSE), the number of model training iterations is set to 1000, and the training accuracy threshold is set to 10⁻⁶.

[0063] The model collects physical field intensity, processing speed, and processing time parameters in real time during the polishing process, predicts the processing quality of each surface sub-region frame by frame, and generates a global quality prediction cloud map.

[0064] Step S7: Collect actual processing quality data through real-time feedback module 7, compare the deviation with the predicted data, and generate a processing error dataset.

[0065] Step S7 specifically includes: The real-time feedback module 7 includes a surface contour detection unit, a roughness detection unit, and a stress detection unit. After a single batch of polishing is completed, the curved surface is measured in its entirety to obtain the actual roughness, material removal amount, and residual stress data.

[0066] Calculate the deviations between the predicted and measured values ​​of the three indicators respectively. The deviation calculation formula is as follows:

[0067] In the formula, Xpre is the predicted quality value, Xreal is the actual detected value, ΔX is the error of a single index, and all regional error data are summarized to form a complete processing error dataset.

[0068] Step S8: The process feature database is iteratively updated based on the error dataset by combining the case reasoning algorithm and the incremental learning algorithm with the self-evolving process library module 8.

[0069] Step S8 specifically includes: The self-evolving process library module 8 contains a massive number of process case samples. It uses a case reasoning algorithm to match historical process cases that are similar to the current surface features and processing conditions, and retrieves the baseline process parameters.

[0070] An incremental learning algorithm is used to input the surface features, process parameters, processing errors, and quality results of this processing into the database as new training samples, and automatically correct the threshold of process parameters for similar working conditions.

[0071] Processing samples with errors exceeding the threshold are marked as key samples, and parameter matching rules are optimized for corresponding curvature, defects, and material conditions to achieve autonomous iterative upgrades of the process library.

[0072] Step S9: The updated process data is called through the cloud-edge collaborative decision-making module 9 to generate the next round of optimized processing strategy, thereby realizing closed-loop iterative polishing processing.

[0073] Step S9 specifically includes: The cloud-edge collaborative decision-making module 9 adopts an edge computing + cloud server architecture. The edge computing nodes are deployed locally on the device and are responsible for real-time process decision-making and dynamic parameter adjustment with a response latency of ≤50ms. The cloud server is responsible for massive data storage, offline model training, and global process optimization.

[0074] Edge nodes retrieve updated process data from the evolutionary process library in real time, and combine it with the current surface features of the part to correct the partition process parameters for the next round of polishing; the cloud summarizes processing data from multiple devices and batches daily, optimizes the weights of reinforcement learning and neural network models, and sends them to local edge nodes to complete model updates.

[0075] The iterative processing error thresholds are set as follows: roughness error ≤ 0.02μm, material removal error ≤ 0.03g, and residual stress error ≤ 5MPa. If the thresholds are met, the processing is deemed qualified; if the thresholds are exceeded, the process parameters are continuously iterated and optimized. Example

[0076] This invention discloses a polishing and grinding device for curved surface parts, comprising: The surface data acquisition module 1 is used to acquire the three-dimensional contour data, surface roughness data, material parameter data, and defect distribution data of the surface part to be processed.

[0077] Digital twin modeling module 2 is used to construct a digital twin model of the part to be processed based on the acquired data, realizing a four-dimensional mapping of geometry, material, defects, and processing status.

[0078] Surface energy field analysis module 3 is used to establish a surface energy distribution model and generate a surface energy field distribution map.

[0079] Process parameter generation module 4 is used to generate differentiated polishing parameters for the corresponding region based on the digital twin model and surface energy field distribution results.

[0080] The multi-physics polishing module 5 is used to complete the surface polishing process by utilizing the synergistic effect of at least two physical fields among electric field, magnetic field, ultrasonic field and plasma field.

[0081] The quality prediction module 6 is used to predict the surface roughness, material removal amount, and residual stress state after polishing in real time.

[0082] The real-time feedback module 7 is used to collect actual processing results and compare them with the predicted results to generate an error dataset.

[0083] The self-evolving process library module 8 is used to update process feature data based on comparison results, enabling autonomous process iteration.

[0084] Cloud-edge collaborative decision-making module 9 is used to call the updated process feature data to generate the next round of optimized processing strategy.

[0085] The process knowledge graph module 10 is used to establish the relationship between part material, curvature characteristics, defect type and process parameters.

[0086] As an optional implementation, the surface data acquisition module 1 of the present invention specifically includes: The surface data acquisition module 1 includes a three-dimensional contour detection unit, a roughness detection unit, a material detection unit, and a defect detection unit. The three-dimensional contour detection unit uses a structured light three-dimensional sensor to collect surface coordinates and curvature parameters. The roughness detection unit uses a laser interferometric roughness sensor. The material detection unit uses an infrared material property sensor. The defect detection unit uses a high-definition industrial vision camera.

[0087] The core parameters collected include local curvature K of the surface, defect density Dd, original surface roughness Ra0, part hardness H, and material density ρ. All data are transmitted to the digital twin modeling module 2 after noise reduction processing, and the data sampling frequency is 10Hz.

[0088] As an optional implementation, the digital twin modeling module 2 of the present invention specifically includes: The digital twin modeling module 2 includes a geometric mapping unit, a material mapping unit, a defect mapping unit, and a processing state mapping unit.

[0089] The geometric mapping unit reconstructs the 3D model based on the surface contour data, with a model resolution of 0.01mm; the material mapping unit accurately assigns values ​​to multiple material parameters; the defect mapping unit completes the visual annotation of defects; and the processing status mapping unit synchronizes the equipment processing parameters and operating status in real time, achieving millisecond-level synchronization between the virtual and real models.

[0090] As an optional implementation, the surface energy field analysis module 3 of the present invention specifically includes: The surface energy field analysis module 3 constructs a curved surface energy field model based on curvature distribution parameters, surface free energy parameters, and defect density parameters. It calculates the surface energy value E of each region through a weighted algorithm, completes the region classification according to the energy value, and generates a high-definition energy field distribution cloud map, providing a quantitative basis for the configuration of differentiated process parameters.

[0091] As an optional implementation, the process parameter generation module 4 of the present invention specifically includes: The process parameter generation module 4 incorporates a reinforcement learning DQN algorithm. With the goal of achieving optimal surface finish, it iteratively optimizes polishing travel speed, physical field strength, processing time, and abrasive parameters. It generates three sets of differentiated process parameter schemes for high, medium, and low energy regions, with a parameter update response speed of ≤100ms.

[0092] As an optional implementation, the multiphysics polishing module 5 of the present invention specifically includes: The multi-physics polishing module 5 includes at least two of the following: magnetorheological processing unit, ultrasonic strengthening processing unit, electrochemical processing unit, and low-temperature plasma processing unit; each unit is integrated into the same polishing actuator, supporting three collaborative processing modes: synchronous, time-sharing, and zone-sharing, adapting to the processing needs of curved parts with different materials and curvatures.

[0093] As an optional implementation, the quality prediction module 6 of the present invention specifically includes: The quality prediction module 6 adopts a three-layer BP neural network prediction model. The input layer consists of an 8-dimensional surface and process parameters, the hidden layer consists of 24 neurons, and the output layer consists of 3-dimensional quality indicators. Through pre-training with a large number of processing samples, the model prediction accuracy error is ≤3%, and it can output the global quality prediction results in real time.

[0094] As an optional implementation, the real-time feedback module 7 of the present invention specifically includes: The real-time feedback module 7 includes a surface profile detection unit, a roughness detection unit, and a stress detection unit. After processing, it automatically completes full-domain detection, compares the deviation between the measured data and the predicted data, classifies and statistically analyzes roughness error, material removal error, and residual stress error, and generates a standardized error dataset.

[0095] As an optional implementation, the self-evolving process library module 8 of the present invention specifically includes: The self-evolving process library module 8 includes a case reasoning submodule and an incremental learning submodule. The case reasoning submodule is used to match similar historical processing cases with benchmark processes; the incremental learning submodule is used to input new processing samples in real time, correct process parameter thresholds, and update model weights to achieve continuous self-evolution and upgrading of the process library.

[0096] As an optional implementation, the cloud-edge collaborative decision-making module 9 of the present invention specifically includes: The cloud-edge collaborative decision-making module 9 uses edge computing nodes to make real-time process decisions and fine-tune parameters to ensure real-time processing. It aggregates global data through cloud servers, completes offline deep training of models and global process optimization, and realizes a two-layer decision-making architecture of real-time edge control and global iteration in the cloud.

[0097] As an optional implementation, the process knowledge graph module 10 of the present invention specifically includes: The process knowledge graph module 10 constructs a multi-dimensional correlation graph of material, curvature, defects, and process parameters, stores a massive number of correlation rules, provides prior knowledge support for the process parameter generation module, and improves the process adaptation efficiency for new working conditions and irregular curved surface parts.

[0098] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A polishing and grinding equipment for curved surface parts, characterized in that, include: Surface data acquisition module (1); Digital twin modeling module (2); Surface energy field analysis module (3); Process parameter generation module (4); Multiphysics polishing module (5); Quality prediction module (6); Real-time feedback module (7); Self-evolving process library module (8); Cloud-edge collaborative decision-making module (9); The surface data acquisition module (1) is connected to the digital twin modeling module (2) and the surface energy field analysis module (3) respectively, and is used to acquire the three-dimensional contour data, surface roughness data, material parameter data and defect distribution data of the surface part to be processed; The digital twin modeling module (2) is used to construct a digital twin model of the curved surface part based on the acquired data; The surface energy field analysis module (3) is connected to the digital twin modeling module (2) and is used to establish a curved surface energy distribution model and generate surface energy field distribution results; The process parameter generation module (4) is connected to the digital twin modeling module (2) and the surface energy field analysis module (3) respectively, and is used to generate polishing process parameters for the corresponding region based on the digital twin model and the surface energy field distribution results; The multiphysics polishing module (5) is connected to the process parameter generation module (4) and is used to perform polishing treatment on curved parts according to the polishing process parameters; The quality prediction module (6) is connected to the multiphysics polishing module (5) and is used to predict the surface roughness, material removal amount and residual stress state after polishing. The real-time feedback module (7) is connected to the multiphysics polishing module (5) and is used to collect actual processing result data; The self-evolving process library module (8) is connected to the quality prediction module (6) and the real-time feedback module (7) respectively, and is used to update the process feature data according to the difference between the prediction result and the actual result; The cloud-edge collaborative decision-making module (9) is connected to the self-evolving process library module (8) and is used to generate optimized processing strategies based on the updated process feature data and feed them back to the process parameter generation module (4).

2. The polishing and grinding equipment for curved parts according to claim 1, characterized in that: The surface data acquisition module (1) includes a three-dimensional contour acquisition unit, a surface roughness detection unit, a material identification unit, and a defect identification unit; The outputs of the three-dimensional contour acquisition unit, surface roughness detection unit, material recognition unit and defect recognition unit are all connected to the digital twin modeling module (2).

3. The polishing and grinding equipment for curved surface parts according to claim 1, characterized in that: The digital twin modeling module (2) includes a geometric mapping unit, a material mapping unit, a defect mapping unit, and a processing state mapping unit; The geometric mapping unit, material mapping unit, defect mapping unit, and machining state mapping unit together construct the digital twin model of the curved part.

4. The polishing and grinding equipment for curved surface parts according to claim 1, characterized in that: The surface energy field analysis module (3) includes a curvature analysis unit, a surface free energy analysis unit, and a defect density analysis unit; The surface energy field model is jointly established by the curvature analysis unit, the surface free energy analysis unit, and the defect density analysis unit.

5. The polishing and grinding equipment for curved surface parts according to claim 1, characterized in that: The process parameter generation module (4) includes a parameter calculation unit, a process matching unit, and a strategy optimization unit; The strategy optimization unit generates polishing parameters based on the digital twin model, surface energy field distribution results, and historical process data.

6. The polishing and grinding equipment for curved parts according to claim 1, characterized in that: The multiphysics polishing module (5) includes at least two of the following: a magnetorheological processing unit, an ultrasonic strengthening processing unit, an electrochemical processing unit, and a low-temperature plasma processing unit.

7. The polishing and grinding equipment for curved surfaces according to claim 1, characterized in that: The quality prediction module (6) includes a surface quality prediction unit, a material removal amount prediction unit, and a residual stress prediction unit. The surface quality prediction unit, material removal amount prediction unit, and residual stress prediction unit are used to output the corresponding prediction results.

8. The polishing and grinding equipment for curved surface parts according to claim 1, characterized in that: The real-time feedback module (7) includes a contour detection unit, a roughness detection unit, and a stress detection unit; The outputs of the contour detection unit, roughness detection unit and stress detection unit are all connected to the self-evolving process library module (8).

9. The polishing and grinding equipment for curved parts according to claim 1, characterized in that: The self-evolving process library module (8) includes a process data storage unit, a case learning unit, and a parameter update unit; The case study unit is used to learn and analyze historical processing data, and the parameter update unit is used to update the process parameter model.

10. The polishing and grinding equipment for curved surface parts according to claim 1, characterized in that: The cloud-edge collaborative decision-making module (9) includes an edge decision-making unit, a cloud training unit, and a policy distribution unit; The edge decision unit is used to generate control strategies in real time, the cloud training unit is used to complete model training and optimization, and the strategy distribution unit is used to send the optimized strategy to the process parameter generation module (4).

Images