Method and system for touch screen production optimization

Abstract

The application provides a touch screen production optimization method, comprising: according to initial distribution characteristics, adopting a preset convolutional neural network model to extract key characteristics of stress distribution and material hardness change, obtaining variation trend data of stress gradient and hardness difference in each region; if each adjustment value in the preliminary adjustment scheme of the cutting parameter exceeds the preset stress gradient threshold, then the genetic algorithm iterative calculation is called through the adaptive control framework to generate final cutting parameter configuration data; comparing the stress distribution dynamic change value and the substrate utilization rate change value in the actual cutting process, if the stress deviation exceeds the standard or the utilization rate is lower than the preset standard, then the feedback adjustment signal data is generated; according to the feedback adjustment signal data, the parameter weight value in the adaptive control framework is adjusted, and combined with the current environmental temperature fluctuation data and the material hardness change data, the stable cutting execution scheme data is generated. The application also provides a system for producing a touch screen by using the above method.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for optimizing the production of touch screens. Background Technology

[0002] Precisely controlling cutting parameters to optimize substrate utilization and stress distribution during substrate cutting is a complex technical challenge. Stress distribution and temperature fluctuation data collected by sensor arrays require high-frequency sampling and convolutional neural network processing to extract key features. However, environmental temperature fluctuations and changes in material hardness affect stress distribution, making accurate adjustment of cutting parameters difficult. While dynamic programming algorithms can calculate initial adjustment schemes, iterative optimization using genetic algorithms is necessary when stress gradient deviations are large. Furthermore, the dynamic changes in stress distribution during actual cutting may deviate from expectations, requiring real-time correction via a path planning module. Maximizing substrate utilization while ensuring cutting quality is a trade-off. Adaptive control frameworks can adjust parameter weights based on feedback, but they cannot completely eliminate uncertainties caused by environmental factors. How to integrate multi-source heterogeneous data to achieve precise control of cutting parameters and continuous optimization of substrate utilization is a pressing technical problem that needs to be solved. Summary of the Invention

[0003] This invention provides a method for optimizing the production of touch screens, mainly including:

[0004] The dynamic data of substrate surface stress distribution and ambient temperature fluctuation data collected by the sensor array are acquired, and high-frequency sampling technology is used to generate initial distribution characteristics including stress value, temperature value and hardness trend.

[0005] Based on the initial distribution characteristics, a pre-defined convolutional neural network model is used to extract key features of stress distribution and material hardness changes, and to obtain data on the changing trends of stress gradient and hardness differences in each region.

[0006] Obtain the current ambient temperature fluctuation value and stress distribution dynamic value, and use dynamic programming algorithm to calculate the preliminary adjustment scheme of cutting parameters with the goal of minimizing stress gradient deviation, including the speed adjustment range, angle adjustment range and depth adjustment range.

[0007] If the adjustment values ​​in the initial adjustment scheme of the cutting parameters exceed the preset stress gradient threshold, the genetic algorithm is called through the adaptive control framework to iteratively calculate and generate the final cutting parameter configuration data.

[0008] Based on the final cutting parameter configuration data, the latest dynamic update value of stress distribution is obtained, and the path planning module is used to correct the stress distribution deviation with the goal of minimizing the stress distribution deviation, thereby generating optimized cutting path distribution data.

[0009] By comparing the dynamic changes in stress distribution during the actual cutting process with the changes in substrate utilization, if there is an excessive stress deviation or a utilization rate lower than the preset standard, feedback adjustment signal data is generated.

[0010] The parameter weights within the adaptive control framework are adjusted based on feedback adjustment signal data, and stable cutting execution scheme data is generated by combining current ambient temperature fluctuation data and material hardness change data.

[0011] Real-time cutting speed stability data and cutting depth control data are collected. The stress distribution dynamic data and utilization rate data are weighted and averaged using data fusion technology to generate substrate utilization rate evaluation results. If the preset target is not met, the stress distribution dynamic data and ambient temperature fluctuation data are updated and recalculated in combination with the previous final cutting parameter configuration data.

[0012] The technical solutions provided by the embodiments of the present invention have the following beneficial effects:

[0013] This invention discloses an adaptive cutting control method based on stress distribution and ambient temperature. The method collects stress distribution and ambient temperature data on the substrate surface using a sensor array, extracts key features using a convolutional neural network, and calculates optimal cutting parameters by combining dynamic programming and genetic algorithms. The invention also includes a path planning module to optimize the cutting path by minimizing stress distribution deviation. During the cutting process, the invention monitors changes in stress distribution and substrate utilization in real time, dynamically adjusting control parameters based on feedback signals. Substrate utilization is evaluated using data fusion technology, and parameters are recalculated if the preset target is not met. This invention effectively addresses changes in material hardness and fluctuations in ambient temperature, achieving a stable and efficient cutting process, significantly improving substrate utilization and cutting quality.

[0014] The present invention also discloses a system for cutting and producing touch screens using the above method. Attached Figure Description

[0015] Figure 1 This is a flowchart of a touch screen production optimization method according to the present invention.

[0016] Figure 2 This is a schematic diagram of a touch screen production optimization method according to the present invention.

[0017] Figure 3 This is another schematic diagram of a touch screen production optimization method according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1-3 This embodiment is a system for producing touch screens, and the optimization method used may specifically include:

[0020] Step S101: Obtain dynamic data of substrate surface stress distribution and ambient temperature fluctuation data collected by the sensor array, and use high-frequency sampling technology to generate initial distribution features including stress value, temperature value and hardness trend.

[0021] Data on substrate surface stress distribution and ambient temperature fluctuations are acquired from a sensor array. Based on the raw data, preprocessing is performed to remove outliers and noise. High-frequency sampling is used to resample the preprocessed data, improving its temporal resolution. Principal component analysis is employed to extract key features of stress and temperature values ​​from the resampled data. Based on these extracted features, a multidimensional feature vector containing stress, temperature, and hardness trends is constructed. Time series analysis is used to model the multidimensional feature vector, capturing dynamic change patterns. Pattern recognition algorithms are then used to identify abnormal patterns and potential risks from the time series model, forming initial distribution characteristics.

[0022] In one embodiment, a sensor array collects stress distribution data on the substrate surface at a sampling frequency of 1000 times per second, while a temperature sensor records ambient temperature fluctuations at a frequency of 100 times per second. The collected stress data is denoised using a Kalman filter algorithm, and the temperature data is smoothed using a moving average method. The processed stress and temperature data are correlated using a multiple linear regression model, yielding a correlation coefficient of 85, indicating that temperature fluctuations have a significant impact on the substrate stress distribution. Based on the processed data, the hardness trend of the substrate surface is calculated using finite element analysis, and a hardness distribution map is obtained through iterative calculation, with a maximum hardness value of 350 HV and a minimum hardness value of 280 HV. The stress, temperature, and hardness trend data are input into a convolutional neural network for feature extraction, generating an initial distribution feature matrix containing stress gradient, temperature change rate, and hardness change rate. The matrix has a dimension of 256×256, and each pixel contains three feature values. Cluster analysis of the feature matrix was used to divide the substrate surface into high-stress, medium-stress, and low-stress regions, with high-stress regions accounting for 15%, medium-stress regions accounting for 60%, and low-stress regions accounting for 25%. The resulting initial distribution features were used for subsequent substrate quality assessment and process optimization decisions.

[0023] Step S102: Based on the initial distribution characteristics, a preset convolutional neural network model is used to extract key features of stress distribution and material hardness changes, and obtain the trend data of stress gradient and hardness difference in each region.

[0024] The initial distribution is processed using a convolutional neural network to extract key patterns in stress distribution. Stress gradients are then separated from these patterns, and region segmentation techniques are employed to generate independent trend data for each region. Temporal smoothing is applied to the trend data to obtain a dynamic sequence of hardness differences. If the dynamic sequence exceeds a preset threshold, an anomaly detection algorithm identifies potential anomaly regions. Based on the distribution characteristics of these anomaly regions, a clustering algorithm is used to determine the correlation patterns between stress gradients and hardness differences. Trend data is updated using these correlation patterns to generate a multidimensional dynamic feature set. Finally, the regional stress distribution features are extracted from this multidimensional dynamic feature set to obtain the global trend of hardness changes.

[0025] In one embodiment, based on the initial distribution feature matrix, a pre-defined convolutional neural network model is used for feature extraction. The network structure includes 5 convolutional layers and 3 fully connected layers, with a 3×3 kernel size, a stride of 1, and ReLU activation. Through forward propagation, key features of stress gradient and hardness change rate are extracted, generating a feature map with a resolution of 128×128. Each feature point contains both stress gradient and hardness change rate values. The feature map is then divided into regions using K-means clustering to classify the substrate surface into high-hardness, medium-hardness, and low-hardness regions. The cluster centers are hardness values ​​of 320HV, 300HV, and 270HV, respectively, with each region accounting for 20%, 50%, and 30% of the total. Further detailed analysis of the high-hardness region is performed, using a Gaussian mixture model to fit the stress gradient distribution. The mean stress gradient in the high-hardness region is found to be 15 MPa / mm, with a variance of 2 MPa / mm. Based on the fitting results, a support vector machine algorithm was used to classify the stress gradient within the high-hardness region, dividing it into stress concentration areas and stress dispersion areas, achieving a classification accuracy of 92%. Correlation analysis was performed between the classification results and the hardness change rate data. The Pearson correlation coefficient showed a correlation coefficient of 78 between stress concentration areas and the hardness change rate, indicating that stress concentration has a significant impact on hardness variation. The final generated key feature data was used for in-depth evaluation and optimization design of the substrate material properties.

[0026] Step S103: Obtain the current ambient temperature fluctuation value and stress distribution dynamic value. Using a dynamic programming algorithm, calculate the preliminary adjustment scheme of cutting parameters with the range of speed adjustment value, angle adjustment value, and depth adjustment value as the goal of minimizing stress gradient deviation.

[0027] Ambient temperature fluctuations and stress distribution dynamics are acquired, and a preliminary numerical set is obtained by collecting real-time data through sensors. A dynamic programming algorithm is used to process this preliminary numerical set to determine an initial scheme of cutting parameters for the speed, angle, and depth adjustment ranges. For this initial cutting parameter scheme, a region partitioning technique is used to separate independent dynamic value sequences of stress distribution. Distribution features are extracted from these independent dynamic value sequences, and a clustering algorithm is used to determine the matching pattern between the speed and angle adjustment values. If the matching pattern exceeds a preset threshold, an anomaly detection algorithm is used to determine the deviation region of the depth adjustment range. Cutting parameters are adjusted based on the deviation region, and a smoothing process is applied to obtain the final parameter set of the adjustment scheme. Environmental data features are extracted from the final parameter set to obtain an optimized adjustment scheme for the speed, angle, and depth adjustment ranges.

[0028] In one embodiment, ambient temperature data is collected in real time using a temperature sensor, and temperature fluctuation values ​​are calculated using a sliding window algorithm with a window size of 10 seconds and a temperature fluctuation range of ±5℃. Simultaneously, a stress sensor is used to acquire dynamic values ​​of stress distribution on the substrate surface, with a sampling frequency of 100Hz, generating a dynamic stress distribution map with a resolution of 256×256. Based on a dynamic programming algorithm, the objective function is defined as the mean square error of the stress gradient, with the initial state being the current stress distribution. The state transition equation represents the relationship between the stress gradient and the cutting parameters, with constraints including a speed adjustment range of 50mm / s to 150mm / s, an angle adjustment range of 30° to 60°, and a depth adjustment range of 5mm to 0mm. Through iterative calculation, the optimal combination of cutting parameters is found to be a speed of 120mm / s, an angle of 45°, and a depth of 2mm, at which the mean square error of the stress gradient is minimized to 8MPa / mm. Further multivariate regression analysis is used to verify the relationship between the cutting parameters and the stress gradient, with regression coefficients of speed -3, angle -2, and depth 4, indicating that depth has the greatest impact on the stress gradient. Based on the regression results, a genetic algorithm was used to optimize the cutting parameters. The population size was 100, the number of iterations was 50, the crossover probability was 8, and the mutation probability was 1. The final optimized cutting parameter combination was a speed of 110 mm / s, an angle of 50°, and a depth of 0 mm. The mean square error of the stress gradient was further reduced to 6 MPa / mm.

[0029] Step S104: If each adjustment value in the preliminary adjustment scheme of cutting parameters exceeds the preset stress gradient threshold, the genetic algorithm is called through the adaptive control framework to iteratively calculate and generate the final cutting parameter configuration data.

[0030] The process begins by acquiring initial configuration data, extracting parameter distribution characteristics, and using statistical tools to calculate the mean and variance of each parameter. The parameter distribution characteristics are then compared to a preset stress gradient threshold. If the threshold is exceeded, the cutting parameter range is adjusted using a regulator to generate updated configuration data. For the updated configuration data, stress gradient change characteristics are extracted, and the rate of change of the stress gradient is calculated using a gradient calculation tool. These stress gradient change characteristics are then compared to preset threshold requirements. If the requirements are not met, the adjustment range is reallocated using a regulator to obtain an optimized parameter distribution. Finally, the optimized parameter distribution is used to generate the final configuration data, and statistical tools are used to update the parameter distribution characteristics, resulting in a stable cutting parameter adjustment scheme.

[0031] Step S105: Based on the final cutting parameter configuration data, obtain the latest dynamic update value of stress distribution, and use the path planning module to correct the stress distribution deviation with the goal of minimizing it, thereby generating optimized cutting path distribution data.

[0032] The stress distribution characteristics of the configuration data are extracted, and the distribution deviation value of the stress distribution characteristics is calculated using SPSS to obtain the initial deviation distribution. Based on the initial deviation distribution, the deviation optimization direction is obtained using the gradient descent method, and the deviation optimization direction is used as the basis for path planning adjustment. The updated distribution data is iteratively calculated using the gradient descent method. If the stress distribution stability characteristics of the updated distribution data meet a preset threshold condition, the final cutting path distribution data is output.

[0033] Step S106: Compare the dynamic change value of stress distribution during the actual cutting process with the change value of substrate utilization. If there is a stress deviation exceeding the standard or the utilization rate is lower than the preset standard, then generate feedback adjustment signal data.

[0034] Stress distribution data during the cutting process is collected by sensors, and dynamic changes in substrate utilization are recorded simultaneously. The stress distribution data and substrate utilization data are input into a pre-established data analysis module to calculate dynamic change characteristic values. These characteristic values ​​are compared with a preset deviation threshold to determine if they exceed the threshold. If they do, a linear regression algorithm is used to fit the dynamic change characteristic values, generating a trend correction signal. The direction of dynamic change in substrate utilization is adjusted based on the trend correction signal. The substrate utilization value is optimized through iterative calculation; after each iteration, the dynamic change characteristic value is recalculated and compared with the deviation threshold until the dynamic change characteristic value meets the preset standard. The optimized substrate utilization value is then input into a dynamic distribution generation module to generate the final optimized substrate utilization distribution.

[0035] Step S107: Adjust the parameter weight values ​​within the adaptive control framework based on the feedback adjustment signal data, and generate stable cutting execution scheme data by combining the current ambient temperature fluctuation data and material hardness change data.

[0036] An initial dynamic dataset is constructed by collecting ambient temperature fluctuation data and material hardness change data using temperature sensors and a hardness tester. The mean, variance, and peak value are extracted from this dataset as characteristic values ​​for the adjustment signal. The PID controller processes these characteristic values ​​to obtain updated parameter weights. The Pearson correlation coefficient between the updated parameter weights and the fluctuation data is calculated. If the correlation coefficient is greater than a preset threshold, a stable cutting trend is obtained. If the stable cutting trend exceeds a preset range, the gradient descent method is used to adjust the weight allocation within the PID controller to obtain optimized weight values. Based on the optimized weight values, stable cutting scheme data, including cutting speed and cutting depth, is generated.

[0037] Step S108: Collect real-time cutting speed stability data and cutting depth control data. Use data fusion technology to weighted average the stress distribution dynamic data and utilization rate data to generate substrate utilization rate evaluation results. If the preset target is not met, update the stress distribution dynamic data and ambient temperature fluctuation data, and recalculate based on the previous final cutting parameter configuration data.

[0038] A weighted average method is used to fuse optimized stress distribution dynamic data, cutting speed data, and cutting depth data to calculate substrate utilization. Based on the calculated substrate utilization, it is determined whether the preset target has been met. If not, a support vector machine algorithm is used to analyze the trends in cutting speed and cutting depth to determine new parameter configuration data. Based on the new parameter configuration data, adjusted stress distribution dynamic data is re-acquired. The adjusted stress distribution dynamic data, substrate utilization data, and temperature fluctuation data are then input into the weighted average method for fusion to obtain the final substrate utilization evaluation result.

[0039] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for optimizing the production of touch screens, characterized in that, The method includes: The dynamic data of substrate surface stress distribution and ambient temperature fluctuation data collected by the sensor array are acquired, and high-frequency sampling technology is used to generate initial distribution characteristics including stress value, temperature value and hardness trend. Based on the initial distribution characteristics, a pre-defined convolutional neural network model is used to extract key features of stress distribution and material hardness changes, and to obtain data on the changing trends of stress gradient and hardness differences in each region. Obtain the current ambient temperature fluctuation value and stress distribution dynamic value, and use dynamic programming algorithm to calculate the preliminary adjustment scheme of cutting parameters with the goal of minimizing stress gradient deviation; If the adjustment values ​​in the initial adjustment scheme of the cutting parameters exceed the preset stress gradient threshold, the genetic algorithm is called through the adaptive control framework to iteratively calculate and generate the final cutting parameter configuration data. Based on the final cutting parameter configuration data, the latest dynamic update value of stress distribution is obtained, and the path planning module is used to correct the stress distribution deviation with the goal of minimizing the stress distribution deviation, thereby generating optimized cutting path distribution data. By comparing the dynamic changes in stress distribution during the actual cutting process with the changes in substrate utilization, if there is an excessive stress deviation or a utilization rate lower than the preset standard, feedback adjustment signal data is generated. The parameter weights within the adaptive control framework are adjusted based on feedback adjustment signal data, and stable cutting execution scheme data is generated by combining current ambient temperature fluctuation data and material hardness change data. Real-time cutting speed stability data and cutting depth control data are collected. The stress distribution dynamic data and utilization rate data are weighted and averaged using data fusion technology to generate substrate utilization rate evaluation results. If the preset target is not met, the stress distribution dynamic data and ambient temperature fluctuation data are updated and recalculated in combination with the previous final cutting parameter configuration data.

2. The method according to claim 1, characterized in that, The process involves acquiring dynamic data on substrate surface stress distribution and ambient temperature fluctuation data collected by the sensor array, and using high-frequency sampling technology to generate initial distribution features including stress values, temperature values, and hardness trends, including: Acquire data on substrate surface stress distribution and ambient temperature fluctuations collected by the sensor array; Based on the collected raw data, data preprocessing is performed to remove outliers and noise; High-frequency sampling technology is used to resample the preprocessed data to improve the temporal resolution of the data. Principal component analysis was used to extract the main features of stress and temperature values ​​from the resampled data. Based on the extracted main features, a multidimensional feature vector containing stress values, temperature values, and hardness trends is constructed. Time series analysis methods are used to perform time-series modeling on multidimensional feature vectors to capture dynamic change patterns; By using pattern recognition algorithms, abnormal patterns and potential risks are identified from time series models, forming initial distribution characteristics.

3. The method according to claim 1, characterized in that, Based on the initial distribution characteristics, a preset convolutional neural network model is used to extract key features of stress distribution and material hardness changes, obtaining trend data on stress gradient and hardness differences in each region, including: The initial distribution is processed in depth using a convolutional neural network to extract key patterns of stress distribution. Stress gradients are extracted from key patterns, and independent trend data for each region are generated using region segmentation techniques. The trend data is smoothed over time to obtain a dynamic sequence of hardness differences. If the dynamic change sequence exceeds the preset threshold, the potential abnormal region is identified by an anomaly detection algorithm; Based on the distribution characteristics of the abnormal areas, a clustering algorithm is used to determine the correlation pattern between stress gradient and hardness difference; Trend data is updated by association patterns to generate a multi-dimensional dynamic feature set; The final regional stress distribution features are extracted from the multidimensional dynamic feature set to obtain the global trend of hardness change.

4. The method according to claim 1, characterized in that, The process involves acquiring current ambient temperature fluctuations and dynamic stress distribution values, and using a dynamic programming algorithm to calculate a preliminary adjustment scheme for cutting parameters, including the ranges for speed, angle, and depth adjustments, with the goal of minimizing stress gradient deviation. The system acquires ambient temperature fluctuations and dynamic stress distribution values, and obtains a preliminary numerical set by collecting real-time data through sensors. The initial cutting parameters are determined by processing the preliminary numerical set using a dynamic programming algorithm, which includes the ranges for speed adjustment, angle adjustment, and depth adjustment. For the initial cutting parameter scheme, a region partitioning technique is used to separate the independent dynamic value sequence of stress distribution; Distribution features are extracted from independent dynamic value sequences, and the matching pattern between speed adjustment values ​​and angle adjustment values ​​is determined by clustering algorithms; If the matching pattern exceeds the preset threshold, the deviation area of ​​the depth adjustment value range is determined by the anomaly detection algorithm; The cutting parameters are adjusted according to the deviation area, and the final parameter set of the adjustment scheme is obtained through smoothing. Environmental data features are extracted from the final parameter set to obtain optimized adjustment schemes for the speed adjustment range, angle adjustment range, and depth adjustment range.

5. The method according to claim 1, characterized in that, If any of the adjustment values ​​in the initial cutting parameter adjustment scheme exceed the preset stress gradient threshold, the genetic algorithm is invoked through an adaptive control framework to iteratively calculate and generate the final cutting parameter configuration data, including: Obtain initial configuration data, extract parameter distribution characteristics from it, and use statistical tools to calculate the mean and variance of each parameter; The parameter distribution characteristics are compared with the preset stress gradient threshold. If the threshold is exceeded, the cutting parameter range is adjusted by the regulator to generate updated configuration data. Based on the updated configuration data, extract the stress gradient change characteristics and use a gradient calculation tool to calculate the rate of change of the stress gradient. The stress gradient change characteristics are compared with the preset threshold requirements. If the requirements are not met, the adjustment value range is redistributed through the regulator to obtain the optimized parameter distribution. The final configuration data is generated based on the optimized parameter distribution. Statistical tools are used to update the parameter distribution characteristics to obtain a stable cutting parameter adjustment scheme.

6. The method according to claim 1, characterized in that, The process involves obtaining the latest dynamic stress distribution value based on the final cutting parameter configuration data, correcting it through the path planning module with the goal of minimizing stress distribution deviation, and generating optimized cutting path distribution data, including: Extract the stress distribution characteristics of the configuration data, and use SPSS to calculate the distribution deviation value of the stress distribution characteristics to obtain the initial deviation distribution; Based on the initial deviation distribution, the deviation optimization direction is obtained using the gradient descent method, and the deviation optimization direction is used as the basis for path planning adjustment. The updated distribution data is calculated iteratively using the gradient descent method. If the stress distribution stability characteristics of the updated distribution data meet the preset threshold condition, the final cutting path distribution data will be output.

7. The method according to claim 1, characterized in that, The dynamic changes in stress distribution during the actual cutting process are compared with the changes in substrate utilization. If the stress deviation exceeds the standard or the utilization rate is lower than the preset standard, feedback adjustment signal data is generated, including: Stress distribution data during the cutting process is collected by sensors, and dynamic changes in substrate utilization are recorded simultaneously. Input the stress distribution data and substrate utilization data into the pre-established data analysis module to calculate the dynamic change characteristic value; The dynamically changing characteristic value is compared with a preset deviation threshold to determine whether it exceeds the threshold. If the threshold is exceeded, a linear regression algorithm is used to fit the dynamically changing feature values ​​and generate a trend correction signal. Adjust the dynamic change direction of substrate utilization based on trend correction signals; The substrate utilization value is optimized by iterative calculation. After each iteration, the dynamic change characteristic value is recalculated and compared with the deviation threshold until the dynamic change characteristic value meets the preset standard. The optimized substrate utilization value is input into the dynamic distribution generation module to generate the final optimized substrate utilization distribution.

8. The method according to claim 1, characterized in that, The step of adjusting the parameter weights within the adaptive control framework based on feedback adjustment signal data, and combining this with current ambient temperature fluctuation data and material hardness change data, generates stable cutting execution scheme data, including: An initial dynamic dataset is constructed by collecting ambient temperature fluctuation data and material hardness change data through temperature sensors and hardness testers. The mean, variance, and peak value are extracted from the initial dynamic dataset as feature values ​​of the modulated signal; The characteristic values ​​of the control signal are processed using a PID controller to obtain updated parameter weights. Calculate the Pearson correlation coefficient between the updated parameter weights and the fluctuation data. If the correlation coefficient is greater than a preset threshold, a stable cutting trend is obtained. If the stable cutting trend exceeds the preset range, the gradient descent method is used to adjust the weight allocation within the PID controller to obtain optimized weight values. Based on the optimized weight values, stable cutting scheme data including cutting speed and cutting depth is generated.

9. The method according to claim 1, characterized in that, The collected real-time cutting speed stability data and cutting depth control data are weighted and averaged using data fusion technology to fuse stress distribution dynamic data and utilization rate data, generating substrate utilization rate evaluation results. If the preset target is not met, the stress distribution dynamic data and ambient temperature fluctuation data are updated, and the results are recalculated based on the previous final cutting parameter configuration data, including: The optimized stress distribution dynamic data, cutting speed and cutting depth data are fused using a weighted average method to calculate the substrate utilization rate. Based on the calculated substrate utilization rate, determine whether the preset target has been achieved; If the desired results are not achieved, the support vector machine algorithm is used to analyze the trends in cutting speed and cutting depth to determine new parameter configuration data. Based on the new parameter configuration data, reacquire the adjusted dynamic data of stress distribution; The adjusted stress distribution dynamic data, substrate utilization data, and temperature fluctuation data are input into a weighted average method for fusion to obtain the final substrate utilization evaluation result data.

Images