Production error control method and system in touch screen production process

Abstract

The invention relates to the technical field of production error control, in particular to a production error control method and system in a touch screen production process. A sensor is deployed in a key process of touch screen production, and process parameters are collected in real time; preprocessing the process parameters to generate preprocessed process parameters and error distribution data; analyzing data by using an error prediction model, predicting an error trend and identifying a potential error accumulation point; using an adaptive parameter adjustment prediction model to analyze the accumulation point and the upstream process parameters, and generating a parameter adjustment coefficient; according to analysis results of the error prediction model and the adaptive parameter adjustment prediction model, a control instruction is generated, production equipment parameters are adjusted in real time, and feedback control is formed; and on the basis of the error data, the parameter adjustment data and the product quality inspection result, model parameters are updated in combination with an optimization algorithm. According to the invention, the accuracy of touch screen production error control can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production error control, and in particular to a production error control method and system for a touch screen production process. Background Art

[0002] Touch screens, as a key component of human-computer interaction, are widely used in smartphones, tablets, industrial control equipment, and other fields. The touch screen production process involves multiple processes (such as thin film deposition, etching, and lamination), and the accuracy of each process directly affects the product's performance and yield. However, existing production processes often face the following problems:

[0003] Error accumulation: Minor errors in multiple processes may accumulate in subsequent processes, resulting in uneven touch screen sensitivity or failure;

[0004] Insufficient adaptive parameter adjustment: Traditional production equipment lacks dynamic error detection and feedback mechanisms, making it difficult to adjust parameters in a timely manner during the production process;

[0005] Excessive intervention by technical personnel: Existing error control relies heavily on work experience, resulting in low efficiency and poor consistency.

[0006] To this end, a production error control method and system for a touch screen production process are proposed. Summary of the Invention

[0007] The purpose of the present invention is to provide a production error control method and system for the touch screen production process. First, sensors are deployed in key touch screen production processes to collect process parameters in real time. The process parameters are preprocessed to generate preprocessed process parameters and error distribution data. Then, an error prediction model is used to analyze the data, predict error trends, and identify potential error accumulation points. An adaptive parameter adjustment prediction model is used to analyze the accumulation points and upstream process parameters to generate parameter adjustment coefficients. Based on the model analysis results, control instructions are generated to adjust production equipment parameters in real time to form feedback control. Based on the error data, parameter adjustment data, and product quality inspection results, the model parameters are updated in combination with an optimization algorithm. The present invention can improve the accuracy and reliability of production error control in industrial control systems.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A production error control method for a touch screen production process, comprising:

[0010] Deploy sensors in key processes of touch screen production to collect process parameters in real time;

[0011] Preprocessing the process parameters to obtain preprocessing process parameters and error distribution data;

[0012] Analyzing the pretreatment process parameters and the error distribution data using an error prediction model to predict error trends and identify potential error accumulation points;

[0013] Analyzing the potential error accumulation point and the process parameters of the upstream process using an adaptive parameter adjustment prediction model to obtain a parameter adjustment coefficient;

[0014] Generate control instructions based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, and adjust the control parameters of the production equipment in real time to form feedback control;

[0015] Update model parameters based on error data, parameter adjustment data, and product quality inspection results, combined with optimization algorithms.

[0016] Furthermore, the process parameters include: substrate transfer parameters, alignment parameters, coating parameters, exposure parameters and lamination parameters.

[0017] Furthermore, the process of preprocessing the process parameters to obtain preprocessed process parameters and error distribution data includes:

[0018] Performing data cleaning on the collected process parameters, removing noise through filtering algorithms, and eliminating outliers;

[0019] Standardizing the process parameters after cleaning to generate the pretreatment process parameters;

[0020] Calculating the error between the pretreatment process parameter and the target value to generate error data;

[0021] Statistical analysis is performed on the error data to fit the error distribution and generate the error distribution data including distribution parameters.

[0022] Furthermore, the process of analyzing the pre-processing parameters and the error distribution data using an error prediction model to predict error trends and identify potential error accumulation points includes:

[0023] Inputting the preprocessing process parameters and the error distribution data into an error prediction model, wherein the error prediction model combines a long short-term memory network, a graph neural network, and a causal reasoning module to construct a causal graph and a topological structure of error transmission between processes;

[0024] Analyzing time series process parameter data through the long short-term memory network to predict the error trend of each process;

[0025] Utilizing the graph neural network and the causal reasoning module, the error transmission relationship between processes is quantified and the probability of occurrence of potential error accumulation points is calculated;

[0026] According to the error trend and the occurrence probability, potential error accumulation points exceeding a preset threshold are identified, and a prediction result is output.

[0027] Furthermore, the process of analyzing the potential error accumulation point and the process parameters of the upstream process using the adaptive parameter adjustment prediction model to obtain the parameter adjustment coefficient includes:

[0028] Inputting potential error accumulation points and pre-processing parameters of upstream processes into an adaptive parameter adjustment prediction model that combines error sensitivity analysis, deep reinforcement learning, and a multi-objective optimization algorithm;

[0029] Through the error sensitivity analysis layer, the influence of upstream process parameters on the potential error accumulation point is quantified to determine the highly sensitive parameters;

[0030] For the highly sensitive parameters, a deep reinforcement learning layer is used to generate preliminary adjustment coefficients based on the error minimization objective;

[0031] Through the multi-objective optimization layer, error control, production efficiency and adjustment cost are balanced to optimize the preliminary adjustment coefficient and generate the parameter adjustment coefficient within the equipment constraint range.

[0032] Furthermore, the process of updating the model parameters based on the error data, parameter adjustment data, and product quality inspection results in combination with the optimization algorithm includes:

[0033] Integrate error data, parameter adjustment data, and product quality inspection results to form a structured data set;

[0034] Analyzing the structured data set using an optimization algorithm, wherein the optimization algorithm includes an adaptive differential evolution algorithm and a multi-objective particle swarm optimization algorithm;

[0035] Using the adaptive differential evolution algorithm to search the parameter space of the error prediction model and the adaptive parameter adjustment prediction model to perform preliminary parameter updates;

[0036] The multi-objective particle swarm optimization algorithm is used to optimize the initially updated parameters based on the multi-objective optimization function, and the optimization weights are adjusted according to the data volatility through the dynamic weight adjustment module to generate updated error prediction model parameters and adaptive parameter adjustment prediction model parameters.

[0037] A production error control system for a touch screen production process comprises: a data acquisition unit, a data processing unit, an error prediction unit, a parameter adjustment prediction unit, an adaptive control unit and a model parameter optimization unit; the data acquisition unit is used to obtain process parameters collected by sensors in real time; the sensors are deployed in key processes of touch screen production; the data processing unit is used to preprocess the process parameters to obtain preprocessed process parameters and error distribution data; the error prediction unit is used to analyze the preprocessed process parameters and the error distribution data using an error prediction model, predict error trends and identify potential error accumulation points; the parameter adjustment prediction unit is used to analyze the potential error accumulation points and the process parameters of the upstream process using an adaptive parameter adjustment prediction model to obtain parameter adjustment coefficients; the adaptive control unit is used to generate control instructions based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, and adjust the control parameters of the production equipment in real time to form feedback control; the model parameter optimization unit is used to update the model parameters based on the error data, parameter adjustment data and product quality inspection results, in combination with an optimization algorithm.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention proposes an error prediction method for predicting error trends and identifying potential error accumulation points; this method combines long-short-term memory networks, graph neural networks and causal reasoning modules, and makes full use of time series data characteristics, relationships between processes and potential causal connections, thereby improving the ability to predict error trends and accumulation, providing more effective data support for subsequent production error control, and further improving the accuracy and reliability of touch screen production error control.

[0040] 2. The present invention proposes a parameter adjustment prediction method for obtaining parameter adjustment coefficients; this method combines error sensitivity analysis, deep reinforcement learning and multi-objective optimization algorithm. Through error sensitivity analysis, it can identify the key process parameters that have the greatest impact on potential error accumulation points to achieve precise error control; through deep reinforcement learning with error minimization as the goal, it automatically explores the best parameter adjustment strategy, and uses a multi-objective optimization algorithm for comprehensive consideration to obtain the global optimal solution, thereby improving the accuracy and reliability of touch screen production error control.

[0041] 3. The present invention proposes a production error control method for adaptive control and model parameter optimization; this method generates control instructions based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, and adjusts the control parameters of the production equipment in real time; based on the error data, parameter adjustment data and product quality inspection results in the control process, and combined with the optimization algorithm, the model parameters are updated, which is conducive to improving the accuracy and reliability of subsequent production error control. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a method for controlling production errors in a touch screen production process according to the present invention;

[0043] Figure 2 Schematic diagram of the structure of the error prediction model of the present invention;

[0044] Figure 3 The present invention is a structural diagram of a production error control system for a touch screen production process. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] See also Figures 1 to 3 The present invention provides a production error control method and system for a touch screen production process. The technical solution is as follows:

[0047] Example 1:

[0048] In order to improve the production quality of touch screens, a touch screen manufacturer uses a production error control method for the touch screen production process proposed by the present invention. The process of this method is shown in the following figure. Figure 1 Shown, including:

[0049] Deploy sensors in key processes of touch screen production to collect process parameters in real time;

[0050] Furthermore, the process parameters are acquired by using different sensor devices;

[0051] Furthermore, the process parameters include: substrate transfer parameters, alignment parameters, coating parameters, exposure parameters and lamination parameters;

[0052] Furthermore, the substrate transmission parameters include transmission speed and transmission position deviation; the alignment parameters include: alignment offset, alignment speed, alignment pressure, etc.; the coating parameters include: coating thickness, coating speed, coating temperature, coating pressure, etc.; the exposure parameters include: exposure amount, exposure time, light source focus, light source wavelength, etc.; the lamination parameters include: bonding pressure, bonding speed, position offset, bubble rate, etc.

[0053] By introducing the process parameters of different production processes, a real and effective data basis is provided for the subsequent prediction of the error correlation and error accumulation of the process parameters of each process, thereby improving the accuracy and reliability of touch screen production error control.

[0054] Preprocessing the process parameters to obtain preprocessing process parameters and error distribution data;

[0055] Furthermore, the process of preprocessing the process parameters to obtain preprocessed process parameters and error distribution data includes:

[0056] Perform data cleaning on the collected process parameters, remove noise through filtering algorithms, and eliminate outliers;

[0057] Standardize the process parameters after cleaning to generate pre-treatment process parameters;

[0058] Calculate the error between the pre-processing parameters and the target value and generate error data;

[0059] Perform statistical analysis on the error data, fit the error distribution, and generate error distribution data including distribution parameters.

[0060] Furthermore, data cleaning includes filtering, outlier detection, and missing value processing. The filtering algorithm uses Kalman filtering, outlier detection uses statistical methods to eliminate abnormal points, and the missing value processing process uses linear interpolation to fill in missing data.

[0061] Furthermore, the standardization process converts the parameters into a distribution with a mean of 0 and a standard deviation of 1. This can unify the parameter formats of different processes and different dimensions, facilitating subsequent cross-process analysis;

[0062] Furthermore, the target values ​​of each parameter will be affected by the production equipment and production environment, so the setting of the target value is not unique;

[0063] Furthermore, the process of generating error distribution data includes: calculating the mean, variance, skewness and other statistical quantities of the error data; modeling the error distribution, fitting it to a normal distribution, and generating distribution parameters; and organizing the distribution parameters according to timestamps to obtain error distribution data.

[0064] By preprocessing the process parameters, data quality can be improved, providing input basis for subsequent cross-process analysis and error prediction, thereby improving the accuracy and reliability of touch screen production error control.

[0065] Analyze pre-processing parameters and error distribution data using error prediction models to predict error trends and identify potential error accumulation points;

[0066] Furthermore, the error prediction model is used to analyze the pre-processing parameters and error distribution data, and the process of predicting error trends and identifying potential error accumulation points includes:

[0067] The pre-processing parameters and error distribution data are input into the error prediction model. The error prediction model combines the long short-term memory network, graph neural network and causal reasoning module to construct a causal graph and topological structure of error transmission between processes.

[0068] Analyze time series process parameter data through long short-term memory networks to predict the error trend of each process;

[0069] Using graph neural networks and causal reasoning modules, we quantify the error transmission relationship between processes and calculate the probability of occurrence of potential error accumulation points.

[0070] Based on the error trend and probability of occurrence, identify potential error accumulation points that exceed the preset threshold and output the prediction results;

[0071] Furthermore, the error prediction model adopts a fusion architecture of hybrid neural network and causal analysis. The structure of the model can be referred to Figure 2 , including: input layer, causal analysis layer, graph neural network layer, long short-term memory network layer, error accumulation judgment layer and output layer;

[0072] Furthermore, the causal analysis layer is used to construct a causal diagram for error transmission between processes. The process is as follows: each process and its corresponding key process parameters are defined as nodes in the causal diagram; an initial causal diagram is constructed based on the process flow and historical experience; constraint-based methods, such as PC and GES algorithms, are used to learn causal relationships from node parameter data to verify and correct the initial causal diagram; finally, for each causal relationship, regression analysis is used to quantify its impact, that is, to quantify the impact of the error in one process on the downstream process. This impact can reflect the intensity of error transmission between processes.

[0073] Furthermore, the graph neural network layer treats processes and corresponding process parameters as nodes and error transmission relationships as edges to model a multi-process topological structure. By combining the graph neural network with the causal graph, the probability of each process combination causing accumulation is obtained, that is, the probability of the occurrence of potential error accumulation points.

[0074] Furthermore, the long short-term memory network layer uses the process parameters and error distribution data of the previous batch of production processes and the current production process to perform time series analysis to obtain the prediction error value of each process;

[0075] Furthermore, the error accumulation judgment layer compares the probability of each process combination causing accumulation with the error accumulation threshold. If it exceeds the threshold, it is marked as a potential accumulation point.

[0076] Three sets of identified potential accumulation point information are given below, denoted as information 1, information 2, and information 3. Each set of information is obtained by analyzing process parameters under different production times and production environments using an error prediction model. The potential accumulation point information includes: process chain, accumulation point probability, and prediction error. The potential accumulation point information is shown in Table 1.

[0077] Table 1. Potential accumulation point information

[0078]

[0079] By combining long-short-term memory networks, graph neural networks, and causal reasoning modules, we can fully utilize the characteristics of time series data, the relationships between processes, and potential causal connections, thereby improving the ability to predict error trends and accumulation, providing more effective data support for subsequent production error control, and further improving the accuracy and reliability of touch screen production error control.

[0080] The adaptive parameter adjustment prediction model is used to analyze the potential error accumulation points and the process parameters of the upstream process to obtain the parameter adjustment coefficient;

[0081] Furthermore, the adaptive parameter adjustment prediction model is used to analyze the potential error accumulation points and the process parameters of the upstream process to obtain the parameter adjustment coefficient. The process includes:

[0082] The potential error accumulation points and pre-processing parameters of the upstream process are input into the adaptive parameter adjustment prediction model, which combines error sensitivity analysis, deep reinforcement learning and multi-objective optimization algorithms;

[0083] Through the error sensitivity analysis layer, the impact of upstream process parameters on potential error accumulation points is quantified to determine highly sensitive parameters;

[0084] For highly sensitive parameters, a deep reinforcement learning layer is used to generate preliminary adjustment coefficients based on the error minimization objective;

[0085] Through the multi-objective optimization layer, the error control, production efficiency and adjustment cost are balanced to optimize the preliminary adjustment coefficient and generate the parameter adjustment coefficient within the equipment constraint range;

[0086] Furthermore, the error sensitivity analysis layer uses an analysis method based on partial derivatives to calculate the partial derivatives of the upstream process parameter error with respect to the parameter error at the potential error accumulation point. The larger the absolute value of the partial derivative, the greater the influence of the upstream process parameter on the potential error accumulation point and the higher the sensitivity. This is used to determine the highly sensitive parameters.

[0087] Furthermore, the deep reinforcement learning layer models the parameter adjustment problem as a Markov decision process, using a reinforcement learning algorithm to train an agent to select appropriate actions based on the current state to maximize the cumulative reward. The reinforcement learning algorithm uses a deep deterministic policy gradient algorithm, where the state is defined as the error value and high-sensitivity parameter value of the potential error accumulation point, the action is defined as the adjustment direction and adjustment amplitude of the high-sensitivity parameter, and the reward is defined as the error value.

[0088] Furthermore, the objective functions of the multi-objective optimization layer are defined as error control, production efficiency, and adjustment cost. Constraints include equipment constraints and process constraints to ensure that process parameters are within a reasonable range. The multi-objective optimization layer adopts a multi-objective optimization algorithm based on genetic algorithm (NSGA-II) to find the Pareto optimal solution set through mechanisms such as non-dominated sorting and congestion distance.

[0089] Based on three sets of potential error accumulation points and upstream process parameter information from different production times and production environments, respectively recorded as sample one, sample two, and sample three, the highly sensitive parameters and corresponding parameter adjustment coefficients of each set of samples were predicted using an adaptive parameter adjustment prediction model. However, in reality, there may be multiple highly sensitive parameters. For ease of presentation, one of them is selected as the reference highly sensitive parameter and reference parameter adjustment coefficient for each set.

[0090] The parameter adjustment prediction test results are shown in Table 2.

[0091] Table 2. Parameter adjustment prediction test results

[0092]

[0093] Error sensitivity analysis can identify the key process parameters that have the greatest impact on potential error accumulation points to achieve precise error control; through deep reinforcement learning with the goal of minimizing errors, the optimal parameter adjustment strategy is automatically explored, and a multi-objective optimization algorithm is used for comprehensive consideration to obtain the globally optimal solution, thereby improving the accuracy and reliability of touch screen production error control.

[0094] Based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, control instructions are generated and the control parameters of the production equipment are adjusted in real time to form feedback control;

[0095] Update model parameters based on error data, parameter adjustment data, and product quality inspection results, combined with optimization algorithms.

[0096] Furthermore, the process of updating the model parameters based on the error data, parameter adjustment data, and product quality inspection results in combination with the optimization algorithm includes:

[0097] Integrate error data, parameter adjustment data, and product quality inspection results to form a structured data set;

[0098] Analyze structured data sets using optimization algorithms, including adaptive differential evolution and multi-objective particle swarm optimization.

[0099] Use the adaptive differential evolution algorithm to search the parameter space of the error prediction model and the adaptive parameter adjustment prediction model to perform preliminary parameter updates;

[0100] Using the multi-objective particle swarm optimization algorithm, the parameters of the initial update are optimized based on the multi-objective optimization function, and the optimization weights are adjusted according to the data fluctuation amplitude through the dynamic weight adjustment module to generate the updated error prediction model parameters and adaptive parameter adjustment prediction model parameters;

[0101] Furthermore, error data, parameter adjustment data, and product quality inspection results are all data collected during the control process;

[0102] Furthermore, the structured data set is a multidimensional data set obtained by associating error data, parameter adjustment data and quality results using timestamps, and includes processes, errors, adjustment records and quality indicators; the adjustment records include parameter values ​​before and after adjustment;

[0103] Furthermore, the adaptive differential evolution algorithm performs preliminary parameter updates by randomly generating multiple sets of model parameters to initialize the population and control parameters; sequentially performing mutation, crossover, and selection operations to iteratively evolve, and automatically adjusting the mutation and crossover rules based on historical performance; after the iterations are complete, selecting the set of parameters with the smallest prediction error as the initial updated model parameters;

[0104] Furthermore, the multi-objective particle swarm optimization algorithm optimizes parameters by: initializing the positions and velocities of multiple particles; each particle represents a set of model parameters, and the particle's initial position is set to the model parameters initially updated by the adaptive differential evolution algorithm; evaluating each particle's performance on the objective function, and adjusting the particle's own speed and position based on its own historical best position and the guidance of the particle with the best objective performance; after the iteration is completed, a set of parameter combinations that perform well on multiple objectives is obtained, from which the parameters that best meet actual needs are selected as the final updated model parameters;

[0105] Furthermore, the objective function in the multi-objective particle swarm optimization algorithm is defined as a weighted combination function of the mean error, defective rate, and production efficiency. The weight coefficients of the mean error, defective rate, and production efficiency in the objective function are adaptively adjusted through a dynamic weight adjustment module according to the change rate of the mean error, the change of the defective rate, and the change rate of production capacity.

[0106] Furthermore, the updated error prediction model parameters and adaptive parameter adjustment prediction model parameters serve the production error control of the next batch.

[0107] By combining the adaptive differential evolution algorithm and the multi-objective particle swarm optimization algorithm, the model parameters can be adaptively adjusted and optimized according to actual production needs, thereby effectively improving the model prediction accuracy, optimizing production parameters and enhancing the stability of the production process.

[0108] This embodiment proposes a production error control method for a touch screen production process. The method first deploys sensors in key touch screen production processes to collect process parameters in real time. The process parameters are preprocessed to generate preprocessed process parameters and error distribution data. An error prediction model is used to analyze the data, predict error trends, and identify potential error accumulation points. An adaptive parameter adjustment prediction model is used to analyze accumulation points and upstream process parameters to generate parameter adjustment coefficients. Based on the model analysis results, control instructions are generated to adjust production equipment parameters in real time, forming feedback control. The model parameters are updated using an optimization algorithm based on the error data, parameter adjustment data, and product quality inspection results. This invention can improve the accuracy and reliability of touch screen production error control.

[0109] Example 2:

[0110] The present invention also proposes a production error control system for a touch screen production process, the structure of the system is as follows: Figure 3 As shown, it includes: a data acquisition unit, a data processing unit, an error prediction unit, a parameter adjustment prediction unit, an adaptive control unit and a model parameter optimization unit;

[0111] The data acquisition unit is used to obtain the process parameters collected by the sensor in real time; the sensor is deployed in the key process of touch screen production;

[0112] The data processing unit is used to pre-process the process parameters to obtain pre-processed process parameters and error distribution data;

[0113] The error prediction unit is used to analyze the pre-processing process parameters and error distribution data using the error prediction model, predict the error trend and identify potential error accumulation points;

[0114] Furthermore, the error prediction model is used to analyze the pre-processing parameters and error distribution data, and the process of predicting error trends and identifying potential error accumulation points includes:

[0115] The pre-processing parameters and error distribution data are input into the error prediction model. The error prediction model combines the long short-term memory network, graph neural network and causal reasoning module to construct a causal graph and topological structure of error transmission between processes.

[0116] Analyze time series process parameter data through long short-term memory networks to predict the error trend of each process;

[0117] Using graph neural networks and causal reasoning modules, we quantify the error transmission relationship between processes and calculate the probability of occurrence of potential error accumulation points.

[0118] Based on the error trend and probability of occurrence, potential error accumulation points exceeding the preset threshold are identified and the prediction results are output.

[0119] The parameter adjustment prediction unit is used to analyze the potential error accumulation points and the process parameters of the upstream process using the adaptive parameter adjustment prediction model to obtain the parameter adjustment coefficient;

[0120] Furthermore, the adaptive parameter adjustment prediction model is used to analyze the potential error accumulation points and the process parameters of the upstream process to obtain the parameter adjustment coefficient. The process includes:

[0121] The potential error accumulation points and pre-processing parameters of the upstream process are input into the adaptive parameter adjustment prediction model, which combines error sensitivity analysis, deep reinforcement learning and multi-objective optimization algorithms;

[0122] Through the error sensitivity analysis layer, the impact of upstream process parameters on potential error accumulation points is quantified to determine highly sensitive parameters;

[0123] For highly sensitive parameters, a deep reinforcement learning layer is used to generate preliminary adjustment coefficients based on the error minimization objective;

[0124] Through the multi-objective optimization layer, error control, production efficiency and adjustment cost are balanced, the preliminary adjustment coefficient is optimized, and the parameter adjustment coefficient is generated within the equipment constraints.

[0125] The adaptive control unit is used to generate control instructions based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, and adjust the control parameters of the production equipment in real time to form feedback control;

[0126] In order to illustrate the effectiveness of the touch screen production error control scheme proposed in the present invention, three different control schemes were used to test the effectiveness of the production error control scheme on randomly sampled test data, and the proportion of error control results of each scheme within a reasonable range was obtained;

[0127] The control schemes are: the control scheme of "error prediction analysis + parameter adjustment prediction" proposed in the present invention, recorded as Scheme 1; removing the error prediction analysis process and predicting the parameter adjustment based on historical experience, recorded as Scheme 2; removing the parameter adjustment prediction process, recorded as Scheme 3; the effectiveness test results of the production error control scheme are shown in Table 3.

[0128] Table 3. Results of the effectiveness test of the production error control scheme

[0129]

[0130] It can be seen from Table 3 that compared with other solutions, the error results obtained by the production error control solution proposed in the present invention are the best, which shows that the strategy of "error prediction analysis + parameter adjustment prediction" is essential for improving the accuracy and reliability of touch screen production error control.

[0131] The model parameter optimization unit is used to update the model parameters according to the error data, parameter adjustment data and product quality inspection results in combination with the optimization algorithm.

[0132] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A production error control method for a touch screen production process, characterized in that: include: Deploy sensors in key processes of touch screen production to collect process parameters in real time; Preprocessing the process parameters to obtain preprocessing process parameters and error distribution data; Analyzing the pretreatment process parameters and the error distribution data using an error prediction model to predict error trends and identify potential error accumulation points; Analyzing the potential error accumulation point and the process parameters of the upstream process using an adaptive parameter adjustment prediction model to obtain a parameter adjustment coefficient; Generate control instructions based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, adjust the control parameters of the production equipment in real time, and form feedback control; Update model parameters based on error data, parameter adjustment data, and product quality inspection results, combined with optimization algorithms.

2. The production error control method of a touch screen production process according to claim 1, characterized in that: The process parameters include: substrate transfer parameters, alignment parameters, coating parameters, exposure parameters and lamination parameters.

3. The production error control method of a touch screen production process according to claim 1, characterized in that: The process of preprocessing the process parameters to obtain preprocessed process parameters and error distribution data includes: Performing data cleaning on the collected process parameters, removing noise through filtering algorithms, and eliminating outliers; Standardizing the process parameters after cleaning to generate the pretreatment process parameters; Calculating the error between the pretreatment process parameter and the target value to generate error data; Statistical analysis is performed on the error data to fit the error distribution and generate the error distribution data including distribution parameters.

4. The method for controlling production errors in a touch screen production process according to claim 1, wherein: The process of analyzing the pre-processing parameters and the error distribution data using an error prediction model, predicting error trends, and identifying potential error accumulation points includes: Inputting the preprocessing process parameters and the error distribution data into an error prediction model, wherein the error prediction model combines a long short-term memory network, a graph neural network, and a causal reasoning module to construct a causal graph of error transmission between processes; Analyzing time series process parameter data through the long short-term memory network to predict the error trend of each process; Utilizing the graph neural network and the causal reasoning module, the error transmission relationship between processes is quantified and the probability of occurrence of potential error accumulation points is calculated; According to the error trend and the occurrence probability, potential error accumulation points exceeding a preset threshold are identified, and a prediction result is output.

5. The production error control method of a touch screen production process according to claim 1, characterized in that: The process of analyzing the potential error accumulation point and the process parameters of the upstream process using the adaptive parameter adjustment prediction model to obtain the parameter adjustment coefficient includes: Inputting potential error accumulation points and pre-processing parameters of upstream processes into an adaptive parameter adjustment prediction model that combines error sensitivity analysis, deep reinforcement learning, and a multi-objective optimization algorithm; Through the error sensitivity analysis layer, the influence of upstream process parameters on the potential error accumulation point is quantified to determine the highly sensitive parameters; For the highly sensitive parameters, a deep reinforcement learning layer is used to generate preliminary adjustment coefficients based on the error minimization objective; Through the multi-objective optimization layer, error control, production efficiency and adjustment cost are balanced to optimize the preliminary adjustment coefficient and generate the parameter adjustment coefficient within the equipment constraint range.

6. The method for controlling production errors in a touch screen production process according to claim 1, wherein: The process of updating model parameters based on error data, parameter adjustment data, and product quality inspection results, combined with optimization algorithms, includes: Integrate error data, parameter adjustment data, and product quality inspection results to form a structured data set; Analyzing the structured data set using an optimization algorithm, wherein the optimization algorithm includes an adaptive differential evolution algorithm and a multi-objective particle swarm optimization algorithm; Using the adaptive differential evolution algorithm to search the parameter space of the error prediction model and the adaptive parameter adjustment prediction model to perform preliminary parameter updates; The multi-objective particle swarm optimization algorithm is used to optimize the initially updated parameters based on the multi-objective optimization function, and the optimization weights are adjusted according to the data fluctuation amplitude through the dynamic weight adjustment module to generate updated error prediction model parameters and adaptive parameter adjustment prediction model parameters.

7. A production error control system for a touch screen production process, characterized in that: include: Data acquisition unit, data processing unit, error prediction unit, parameter adjustment prediction unit, adaptive control unit and model parameter optimization unit; The data acquisition unit is used to obtain process parameters collected by the sensor in real time; Sensors are deployed in key processes of touch screen production; The data processing unit is used to preprocess the process parameters to obtain preprocessed process parameters and error distribution data; the error prediction unit is used to use the error prediction model to analyze the preprocessed process parameters and the error distribution data, predict error trends and identify potential error accumulation points; the parameter adjustment prediction unit is used to use the adaptive parameter adjustment prediction model to analyze the potential error accumulation points and the process parameters of the upstream process to obtain parameter adjustment coefficients; the adaptive control unit is used to generate control instructions based on the analysis results of the error prediction model and the adaptive parameter adjustment prediction model, and adjust the control parameters of the production equipment in real time to form feedback control; the model parameter optimization unit is used to update the model parameters based on error data, parameter adjustment data and product quality inspection results, and in combination with the optimization algorithm.

Images