A polishing-etching integrated processing technology of high-finish topaz material

Abstract

The application discloses a polishing-etching integrated processing technology for high-finish jade crystal stone material, and the technology comprises the following steps: providing an integrated processing device integrating chemical mechanical polishing and chemical etching functions, and deploying a multi-source sensor online monitoring system; synchronously collecting multi-dimensional parameters such as polishing pressure, torque, temperature, pH value, fluoride ion concentration, acoustic emission signal and online roughness; adopting a random forest algorithm to screen key features, and establishing a surface roughness real-time prediction model based on support vector regression or artificial neural network; the application realizes real-time sensing, accurate prediction and self-adaptive regulation of jade crystal stone polishing-etching integrated processing, the processing efficiency is improved by more than 50%, the surface roughness is less than or equal to 0.01 mu m, the good product rate is greater than or equal to 94%, and the application is suitable for high-finish jade crystal stone material processing in the fields of high-end building decoration, electronic cover plate and precise optical device.

Description

Technical Field

[0001] This invention relates to the field of inorganic non-metallic material surface processing technology, and in particular to an integrated polishing-etching process for high-gloss jade crystal materials. Background Technology

[0002] Microcrystalline glass (also known as crystalline glass) is a composite material formed by controlling the heat treatment of base glass to precipitate microcrystals. It is widely used in high-end architectural decoration, electronic product cover plates, precision optical components, and laser gyroscope optical cavities. With increasingly stringent surface quality requirements in various applications, the surface smoothness of microcrystalline glass has become a core indicator of product quality. For example, military and high-quality optical components require a surface roughness of less than 0.1 nanometers for microcrystalline glass, along with strict requirements for surface shape, defects, and cleanliness.

[0003] Current high-gloss surface finishing processes for jade crystal typically employ a discrete, multi-step approach: mechanical grinding → rough mechanical polishing → chemical etching → fine mechanical polishing. This approach suffers from the following technical shortcomings:

[0004] (1) Separation of processes leads to low processing efficiency. Polishing and etching are carried out on different equipment, at different stations and at different times. Multiple cleaning, drying and transfer are required between processes, which not only prolongs the production cycle, but also increases equipment investment and labor costs.

[0005] (2) Lack of real-time sensing methods in the processing. In existing processes, the monitoring of processing status (polishing pressure, temperature, etching solution concentration, material removal rate, surface roughness, etc.) mainly relies on offline detection—that is, the workpiece is taken out of the equipment and measured after processing. This offline detection method not only leads to batch-to-batch feedback lag, but more seriously, it cannot intervene in abnormal states during the processing in a timely manner.

[0006] (3) The processing status is difficult to predict accurately. The integrated polishing-etching process involves a high degree of coupling between chemical etching reaction kinetics and mechanical grinding removal mechanism. More than a dozen parameters affect the processing quality, and the interaction between these parameters is highly nonlinear. Existing research mainly focuses on offline prediction of single processing steps, and a dynamic prediction model integrating online real-time sensing data has not yet been established for the integrated polishing-etching process.

[0007] (4) Process parameter control relies on experience and lacks self-adaptive capability. Currently, the parameter settings for integrated polishing-etching processing mainly rely on the operator's experience and offline experiments. When the processing environment or workpiece condition changes, the preset fixed parameters cannot be automatically adjusted, resulting in large fluctuations in processing quality and unstable yield.

[0008] Therefore, developing an intelligent polishing-etching integrated processing technology that combines online multi-source sensor monitoring, real-time surface quality prediction, and adaptive parameter control to achieve real-time perception, accurate prediction, and dynamic control of the entire processing of jade crystal materials has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0009] The purpose of this invention is to provide an integrated polishing-etching process for high-gloss jade crystal materials, which solves the technical problems of low processing efficiency, lack of real-time perception of the processing process, difficulty in accurately predicting the processing status, and reliance on experience for process parameter control caused by the separation of polishing and etching processes in the prior art. This invention enables real-time monitoring of multiple parameters, dynamic prediction of surface quality, and adaptive control of processing parameters in the integrated polishing-etching process of jade crystal materials, significantly improving processing efficiency, surface quality, and batch consistency.

[0010] To achieve the above objectives, the present invention provides the following technical solution: an integrated polishing-etching process for high-gloss jade crystal materials, comprising the following core technical contents:

[0011] I. Integrated Polishing-Etching Processing Equipment and Online Monitoring System

[0012] This device integrates chemical mechanical polishing and chemical etching functions at the same workstation. The processing device includes an upper polishing disc (which can apply adjustable polishing pressure and has a polyurethane polishing pad attached to its surface), a lower polishing disc (which can adjust the rotation speed and has a non-woven polishing pad attached to its surface), a workpiece fixture (for fixing the jade crystal workpiece), a processing fluid supply system (for continuously supplying integrated processing fluid between the polishing pad and the workpiece), and a temperature control system (for controlling the processing fluid temperature within the range of 25℃ to 45℃).

[0013] Multi-source sensors are deployed at key locations of the processing equipment to form an online monitoring system, including:

[0014] Pressure sensor (installed on the upper polishing disc drive cylinder to monitor polishing pressure);

[0015] Torque sensor (installed on the output shaft of the lower polishing disc drive motor to monitor the drive motor torque);

[0016] Thermocouples (installed in the processing fluid reservoir and workpiece fixture to monitor the processing fluid temperature and workpiece temperature);

[0017] pH electrode (installed in the bypass of the processing fluid circulation pipeline to monitor the pH value of the processing fluid).

[0018] Fluoride ion selective electrode (installed in the bypass of the processing fluid circulation line to monitor fluoride ion concentration).

[0019] Acoustic emission sensor (mounted on the lower polishing disc substrate or workpiece fixture to monitor acoustic emission signals during the machining process);

[0020] Online roughness measurement module (installed at an independent measurement station, using the principle of white light interference or laser confocalization to quickly measure the surface roughness of the workpiece during processing pauses);

[0021] The sampling frequencies of each sensor are as follows: pressure sensor and torque sensor 10~100Hz, temperature sensor 1~10Hz, pH electrode and fluoride ion selective electrode 0.1~1Hz, acoustic emission sensor 100~1000Hz, and online roughness measurement module measures 1~3 times per batch.

[0022] II. Multi-source sensor signal acquisition and preprocessing

[0023] The output signals of the aforementioned multi-source sensors are synchronously acquired at a fixed sampling frequency. The sensor signals acquired at time t are used to form the original feature vector. The acquired original sensor signals are preprocessed, including: using a low-pass filter to remove high-frequency noise, using a moving average method to smooth the signal, and using a min-max normalization method to uniformly map each feature dimension to the [0,1] interval to eliminate the influence of the difference in the dimensions of different sensors on model training.

[0024] III. Real-time Surface Roughness Prediction Based on Hybrid Machine Learning Model

[0025] (1) Feature Importance Analysis and Feature Selection: The Random Forest algorithm was used to evaluate the feature importance of the preprocessed sensor signals. The Random Forest quantifies the contribution of each feature to the prediction target (surface roughness) by calculating the average reduction in impurity (the reduction in Gini coefficient or the reduction in mean square error) across all decision trees. The top k features with a cumulative contribution rate of 95% were selected as the input feature subset of the final prediction model, sorted from largest to smallest importance score. The number of decision trees in the Random Forest was between 100 and 500.

[0026] (2) Support Vector Regression Prediction Model (Small Sample Stage): In the early stages of process development, the historical data sample size is limited (usually less than 200 sets). Support Vector Regression (SVR) is used to establish a surface roughness prediction model. SVR maps the input features to a high-dimensional space through a kernel function, and searches for the optimal linear regression function in the high-dimensional space. The kernel function used is the radial basis function (RBF kernel). The kernel function parameter γ and the regularization parameter C are optimized and selected through grid search combined with cross-validation. The output of the SVR model is the predicted surface roughness. The model exhibits good generalization performance under small sample conditions.

[0027] (3) Deep Prediction Model Using Artificial Neural Networks (Big Data Stage): With the accumulation of production data (historical data sample size exceeding 1000 sets), a deep prediction model is constructed using artificial neural networks (ANNs). The ANN model adopts a three- or four-layer structure: the number of nodes in the input layer equals the number of filtered sensor features, the hidden layers use 2 to 3 layers, the number of nodes in each layer decreases progressively, and the number of nodes in the output layer is 1. The Adam optimizer is used for parameter updates, and Dropout technology (dropout rate 0.1 to 0.3) is used to prevent overfitting. On the test set, the prediction accuracy of the deep ANN model can reach an average absolute percentage error of less than 10%.

[0028] (4) Fusion strategy of hybrid model: In practical application, a phased model deployment strategy is adopted: when the amount of historical data is less than 200 sets, the SVR model is used as the main prediction tool; when the amount of data reaches 200~1000 sets, the weighted fusion model of SVR and ANN is used; when the amount of data exceeds 1000 sets, the ANN model is used as the main prediction tool.

[0029] (5) Rapid cross-formulation adaptation based on transfer learning: When the formulation of jade crystal material or the formulation of processing fluid changes, the prediction accuracy of the original prediction model will decrease significantly under the new conditions. To solve this problem, transfer learning technology is introduced: a deep ANN model trained under the original formulation conditions is used as the "source model"; the first few layers of the source model are frozen (these layers learn the basic mapping relationship from general sensor signals to surface quality); the last 1 to 2 layers of the source model are fine-tuned only on a small number of samples (30 to 50 groups) under the new formulation conditions; thus achieving rapid adaptation of the prediction model under the new formulation conditions.

[0030] IV. Process State Inversion and Dynamic Estimation of Material Removal Rate Based on Online Torque Signals

[0031] During the integrated polishing-etching process, the torque output by the lower polishing disc drive motor... The material removal rate is directly related to the total frictional force on the workpiece surface, and thus has a positive correlation with the material removal rate. Drawing on the basic idea of ​​the Preston equation, the material removal rate can be expressed as... ,in The material removal efficiency coefficient. For polishing pressure, The relative velocity. This invention uses online torque signals to... Dynamic estimation is performed to establish a torque-material removal rate mapping model: , where α and β are model coefficients determined through offline calibration experiments.

[0032] Calibration experimental method: Under fixed processing conditions, process standard-sized jade crystal workpieces and record the torque signals during the processing. After processing, the actual material removal thickness is measured by weighing or thickness measurement, the average material removal rate is calculated, and α and β are determined by linear regression.

[0033] During actual processing, the system dynamically estimates the current material removal rate based on real-time torque signals. .when When the value is significantly lower than expected (e.g., below 80% of the expected value), it indicates that the etching components in the processing fluid have been significantly consumed or the polishing pad has been severely worn. The system will automatically trigger the replenishment of processing fluid or issue a warning to replace the polishing pad.

[0034] V. Adaptive Parameter Control Strategy Based on Multi-Objective Optimization

[0035] (1) Construction of multi-objective optimization function: based on the predicted surface roughness Minimize and material removal rate Maximizing the dual objectives, constructing a multi-objective optimization function. ,in For the target surface roughness, The target material removal rate, and The weighting coefficients are satisfied. .

[0036] (2) Adjustable parameters and control actions: The adjustable control parameters of the system include: polishing pressure Polishing disc rotation speed Processing fluid supply rate Processing fluid temperature and the fluoride ion concentration adjusted by the replenishment system. and abrasive solid content When the predicted processing quality deviates from the target range (i.e., J exceeds the preset threshold) When this happens, the system will automatically trigger the following control actions:

[0037] If R and Increase polishing pressure and polishing disc speed, and check if the fluoride ion concentration needs to be replenished.

[0038] like and Normal: Prioritize increasing the processing fluid supply rate or increasing the fluoride ion concentration;

[0039] like Normal and Prioritize increasing polishing pressure or polishing disc speed;

[0040] like and Reduce polishing pressure and polishing disc speed.

[0041] (3) Decoupling strategy for multivariable coordinated regulation: Since there is a coupling effect between multiple control parameters, the system adopts a decoupling strategy based on model predictive control: using a pre-trained machine learning prediction model, different candidate parameter combinations are simulated and predicted before each regulation, and the parameter combination that makes the prediction comprehensive objective function J optimal is selected for execution.

[0042] VI. Intelligent determination of processing endpoint

[0043] The system automatically determines the processing endpoint when both of the following conditions are met:

[0044] Condition 1: Within N consecutive prediction periods (N ranges from 3 to 5), the predicted surface roughness... Stabilize at the target surface roughness It is within 1.2 times that of the previous period, and there is no obvious downward trend.

[0045] Condition 2: Predicted material removal rate Below the preset threshold (This indicates that the processing has reached a plateau, and further processing will have limited effect on improving surface quality.)

[0046] Once the endpoint conditions are met, the system automatically issues a stop command and stores all sensor data and prediction results from the current processing to the process database for continuous optimization of the subsequent model.

[0047] VII. Cleaning and Drying After Processing

[0048] After the integrated processing is completed, the following cleaning process is adopted: the workpiece surface is rinsed with high-pressure deionized water for 30-60 seconds to remove most of the residual processing fluid; the workpiece is immersed in deionized water and cleaned with 0.8-1.2MHz megasonic waves for 2-5 minutes to remove submicron particles using the micro-jet effect of megasonic waves; ultrasonic cleaning is performed with 40-80kHz for 3-8 minutes; finally, the workpiece is dried with clean hot air (50-70℃) for 3-5 minutes to obtain the finished product.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) Significantly improves processing efficiency: The integrated polishing-etching process eliminates the cleaning, transfer and waiting time between processes, reducing the total processing time by more than 50%. Adaptive parameter control keeps the processing in an optimal state at all times, increasing the yield rate from 78% to over 94%.

[0051] (2) Significantly improve surface quality: Based on online monitoring by multi-source sensors and real-time prediction by hybrid machine learning, the system can promptly detect deviations in processing status and make adaptive adjustments. The surface roughness Ra can be reduced to below 0.01μm, achieving the requirements of mirror-level smoothness.

[0052] (3) Shorten the process development cycle: Based on the rapid cross-formula adaptation technology of transfer learning, when the jade crystal formula changes, only 30 to 50 new samples are needed to complete the fine-tuning and adaptation of the prediction model, and the development cycle is shortened by more than 90%.

[0053] (4) Reduce production costs: Through real-time forecasting and precise control, the consumption of processing fluid is reduced by about 30%, which reduces the use of chemicals and waste emissions; there is no need for inter-process cleaning equipment, and the equipment investment and labor costs are significantly reduced.

[0054] (5) Realize the digitalization and intelligentization of the processing: The system automatically records all sensor data and prediction results to form a traceable process database, providing data support for continuous process optimization. Attached Figure Description

[0055] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0056] Figure 1 This is a flowchart of the integrated polishing-etching process for the high-gloss jade crystal material of the present invention.

[0057] Figure 2 This is a flowchart illustrating the adaptive parameter control strategy of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Please see Figures 1 to 2 The technical solution provided by this invention:

[0060] Example 1

[0061] (1) System setup and sensor deployment

[0062] A multi-source sensor system was deployed on a double-sided polishing machine: a pressure sensor (range 0~0.5 kg / cm², accuracy ±0.01 kg / cm²) was installed at the cylinder output end; a torque sensor (range 0~20 N·m, accuracy ±0.1 N·m) was installed on the output shaft of the lower polishing disc drive motor; K-type thermocouples were installed in the machining fluid reservoir and workpiece fixture, respectively; a pH electrode and a fluoride ion selective electrode (measurement range 0.1~5.0 mol / L) were installed in the bypass of the machining fluid circulation pipeline; a broadband acoustic emission sensor (response frequency 100~1000 kHz) was installed on the lower polishing disc substrate; and a white light interferometric online roughness measurement module (measurement range 0.001~1.0 μm, resolution 0.001 μm) was configured at an independent measurement station. The sampling frequencies of each sensor were: pressure / torque 100 Hz, temperature 10 Hz, pH / F⁻ concentration 0.5 Hz, acoustic emission 500 Hz, and offline roughness was measured once per batch.

[0063] (2) Preparation of processing fluid

[0064] Prepare the polishing-etching integrated composite processing fluid (total volume 1000mL) according to the following formula: ammonium fluoride (NH4F) 76g, nano CeO2 abrasive 60g (average particle size 80nm), citric acid 30g, sodium citrate 15g, sodium polyacrylate 3g, fatty alcohol polyoxyethylene ether 1g, and deionized water to a final volume of 1000mL. The pH of the processing fluid was measured to be 5.2.

[0065] (3) Data acquisition and model training

[0066] Jadeite workpieces were processed under the following initial conditions (polishing pressure 0.12 kg / cm², polishing disc speed 80 rpm, processing fluid supply rate 40 mL / min, processing fluid temperature 35℃). Fifty batches were processed continuously, with each batch taking 60 minutes to process. Surface roughness was measured every 15 minutes. All sensor signals were collected simultaneously, yielding approximately 200 sets of valid data samples.

[0067] A random forest algorithm (200 decision trees) was used to perform feature importance analysis on the training set, selecting the top 6 features that contributed most to surface roughness prediction: polishing pressure, torque, fluoride ion concentration, processing fluid temperature, processing time, and root mean square value of acoustic emission signal. Using these 6 selected features as input, a prediction model was built using support vector regression (SVR). A radial basis function was used as the kernel function, and the optimal parameters γ=0.5 and C=10.0 were determined through grid search. On the test set, the SVR model achieved a mean absolute percentage error of 18.7%.

[0068] (4) Application of online prediction and adaptive regulation

[0069] In subsequent processing, the system collects sensor signals in real time and predicts the surface roughness under the current processing conditions every 30 seconds using a trained SVR model. During one processing cycle, the system predicts... It continued to rise, approaching 0.025 μm, more than twice the target value of 0.01 μm, while... The flow rate was below 80% of the target value of 3 μm / min. The system determined this to be due to insufficient chemical activity in the processing fluid and automatically implemented the following adjustments: increasing the processing fluid supply rate from 40 mL / min to 60 mL / min, increasing the processing fluid temperature from 35°C to 40°C, and triggering the processing fluid replenishment system to add ammonium fluoride solution. Ten minutes after these adjustments were implemented, It decreased to 0.018 μm. The speed was restored to 2.8 μm / min. Processing continued until the 60-minute endpoint, with a measured surface roughness Ra = 0.012 μm and a material removal rate of 42 μm, meeting the processing quality requirements.

[0070] (5) Cleaning and drying

[0071] After processing, the product is rinsed with high-pressure deionized water for 45 seconds, cleaned with megaphonic sound for 3 minutes, cleaned with ultrasonic sound for 5 minutes, and dried with hot air for 4 minutes to obtain the finished product.

[0072] Example 2

[0073] (1) Data accumulation and model upgrading

[0074] After processing 200 batches, the historical data sample size reached approximately 1000 groups. A three-layer artificial neural network (6 nodes in the input layer, 64 nodes in the first hidden layer, 32 nodes in the second hidden layer, and 1 node in the output layer) was retrained to create a deep prediction model. The ReLU activation function, Adam optimizer, learning rate of 0.001, dropout rate of 0.2, and 200 training epochs were used. On the test set, the deep ANN model achieved a mean absolute percentage error (MAPE) of 8.3%. By weighting and ensembling the two models (SVR weight 0.2, ANN weight 0.8), the MAPE of the ensemble model decreased to 7.5%.

[0075] (2) Rapid adaptation of new formula

[0076] Adjusting the amount of nanoparticles in the jade crystal formula (increasing the amount of nano-TiO2 by 2%) altered the chemical etching behavior of the processing solution, causing the original prediction model's accuracy to drop to MAPE = 22.5% under the new conditions. Using a deep ANN model trained under the original formula as the source model, the first two layers (from the input layer to the first hidden layer) were frozen, and the third and output layers were fine-tuned only on 50 samples under the new formula conditions (20 training epochs, learning rate 0.0001). After fine-tuning, the new model's MAPE on the test set recovered to 9.8%.

[0077] (3) Processing and cleaning (same as in Example 1)

[0078] Performance comparison results:

[0079] Total processing time (min) 60 55 120 Surface roughness Ra (pm) 0.012 0.009 0.018 Yield rate (%) 94% 97% 78% Processing liquid consumption (L / batch) 2.4 2.2 3.5 Number of cleaning between processes 0 0 3

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A polishing-etching integrated processing technology for high-gloss jade crystal materials is characterized by, Includes the following steps: (1) Provide an integrated polishing-etching processing device, wherein the device integrates chemical mechanical polishing and chemical etching functions at the same workstation, including an upper polishing disc, a lower polishing disc, a workpiece fixture, a processing fluid supply system and a temperature control system; (2) A multi-source sensor online monitoring system is deployed on the processing device. The multi-source sensors include: a polishing pressure sensor, a polishing disc drive motor torque sensor, a processing fluid temperature sensor, a workpiece temperature sensor, a processing fluid pH electrode, a fluoride ion selective electrode, an acoustic emission sensor, and an online roughness measurement module. (3) The output signals of each sensor are collected synchronously at a preset sampling frequency, and the original signals are preprocessed by denoising, filtering and normalization to construct a multi-dimensional processing state feature vector; (4) A real-time surface roughness prediction model is established using a hybrid machine learning method. The pre-processed sensor signal is used as input to predict the surface roughness of the jade crystal workpiece during the processing. ; (5) Based on the torque signal of the polishing disc drive motor, a dynamic estimation model for material removal rate is established to evaluate the material removal rate in real time during the processing. ; (6) With the predicted surface roughness minimization and the estimated material removal rate maximization as dual objectives, a multi-objective optimization function is constructed. When the predicted processing quality deviates from the target range, one or more of the following are automatically adjusted: polishing pressure, polishing disc rotation speed, processing fluid supply rate, processing fluid temperature, and processing fluid composition concentration. (7) When the predicted surface roughness is stable below the target threshold and the material removal rate is lower than the preset threshold, determine the processing endpoint and stop the machine; (8) After processing, the workpiece is cleaned and dried to obtain a high-gloss jade crystal finished product.

2. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, The sampling frequencies of the multi-source sensors in step (2) are: 10~100Hz for pressure sensor and torque sensor, 1~10Hz for temperature sensor, 0.1~1Hz for pH electrode and fluoride ion selective electrode, and 100~1000Hz for acoustic emission sensor; the online roughness measurement module adopts the principle of white light interference or laser confocalization to measure the surface roughness of the workpiece during the processing pause.

3. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, The hybrid machine learning method in step (4) includes: using the random forest algorithm to evaluate the feature importance of the preprocessed sensor signal, and selecting the top k features with a cumulative contribution rate of 95% as the input feature subset of the prediction model; the number of decision trees in the random forest algorithm is 100 to 500.

4. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, The surface roughness real-time prediction model described in step (4) adopts a support vector regression model when the historical data sample size is less than 200 sets, and the kernel function adopts a radial basis function; when the historical data sample size is greater than 200 sets, an artificial neural network model is adopted. The artificial neural network model contains a three- or four-layer structure, the number of input layer nodes is equal to the number of sensor features after screening, the hidden layer adopts 2 to 3 layers, and the number of output layer nodes is 1.

5. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, When the formulation of jade crystal material or processing fluid changes, a transfer learning method is adopted: a deep artificial neural network model trained under the original formulation conditions is used as the source model, the first few layers of the source model are frozen, and the last 1 to 2 layers of the source model are fine-tuned only on a small number of samples under the new formulation conditions, so as to achieve rapid adaptation of the prediction model under the new formulation conditions.

6. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, The dynamic estimation model for material removal rate mentioned in step (5) is as follows: ,in The torque of the polishing disc drive motor is given by α and β, which are model coefficients determined through offline calibration experiments; when When the level falls below 80% of the expected value, an automatic warning will be triggered to replenish the processing fluid or replace the polishing pad.

7. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, The multi-objective optimization function mentioned in step (6) is: ,in For the predicted surface roughness, For the target surface roughness, For the predicted material removal rate, The target material removal rate, and The weighting coefficients are and satisfy the following conditions: ; When J exceeds the preset threshold, the following control actions are automatically triggered: like and At the same time, increase the polishing pressure and the polishing disc speed; like and Under normal circumstances, increase the processing fluid supply rate or increase the fluoride ion concentration; like Normal and At this time, increase the polishing pressure or the polishing disc speed; like and At this time, reduce the polishing pressure and the polishing disc speed.

8. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, In step (6), a decoupling strategy based on model predictive control is adopted. The trained machine learning prediction model is used to perform simulation prediction on different candidate parameter combinations, and the parameter combination that makes the prediction comprehensive objective function J optimal is selected for execution.

9. The integrated polishing-etching process for high-gloss jade crystal materials according to claim 1, characterized in that, The conditions for determining the processing endpoint in step (7) are: within 3 to 5 consecutive prediction cycles, the predicted surface roughness is stable within 1.2 times the target surface roughness and there is no obvious downward trend, while the predicted material removal rate is lower than the preset threshold.

10. A high-gloss jade crystal material prepared by an integrated polishing-etching process according to any one of claims 1 to 9, characterized in that, The surface roughness of the jade crystal material The surface is free of polishing residue and etching pits.

Images