Method for optimizing TFT (Thin Film Transistor) material manufacturing process

Abstract

The invention relates to the technical field of TFT material manufacturing process data processing, provides a method for optimizing a TFT material manufacturing process, solves the problem that foreign matter and defects cannot be effectively controlled in the TFT material manufacturing process, and obtains an equipment control model by obtaining TFT material manufacturing historical data and utilizing random forest model training. And outputting the optimized control parameters. In the manufacturing process of the TFT material, the image processing model is adopted for foreign matter and defect detection, the control parameters are adjusted according to the detection result, if adjustment is not successful, the control model is further optimized through the genetic algorithm until the qualified TFT material is obtained, the quality and production efficiency of the TFT material are effectively improved, and the production cost is reduced. The rejection rate and the operation cost are reduced, and powerful support is provided for the technical progress of the TFT material manufacturing industry.

Description

Technical Field

[0001] The present invention relates to the technical field of TFT material manufacturing process data processing, and in particular to a method for optimizing TFT material manufacturing process. Background Art

[0002] The manufacturing process of TFT (Thin Film Transistor) materials involves multiple steps. First, a gate wiring pattern is formed on a borosilicate glass substrate by processes such as sputtering, mask exposure, development, and dry etching. PECVD (plasma enhanced chemical vapor deposition) is used for continuous film formation to form a multilayer structure including SiNx film, non-doped a-Si film, and phosphorus-doped n+a-Si film, and the a-Si pattern of the TFT part is formed again by mask exposure and dry etching. Subsequently, a transparent electrode (ITO film) is formed by sputtering film formation, and a display electrode pattern is formed by mask exposure and wet etching. The contact hole pattern, source, drain, and signal line pattern of the gate end insulating film are formed by a preset process, and a protective insulating film is formed by PECVD. Finally, a color filter pattern and an ITO conductive layer are formed on the color filter substrate, combined with the TFT substrate, and liquid crystal material is poured to complete the production of the liquid crystal box.

[0003] In the process of data processing in the TFT (Thin Film Transistor) material manufacturing process, the following technical pain points exist: the TFT manufacturing process is highly dependent on manufacturing equipment, including sputtering machines, PECVD equipment, photolithography machines, etc. In the TFT manufacturing process, it is necessary to monitor and process the process parameters in real time, including parameters such as film thickness, photolithography accuracy, and etching rate. Foreign matter and defects are common problems in the TFT manufacturing process. Foreign matter and defects come from raw materials, equipment, environment, etc. Foreign matter and defects will affect the performance of TFT and even cause the product to be scrapped. The existing TFT material manufacturing process cannot effectively control the generation of foreign matter and defects during the manufacturing process. In order to solve this technical pain point, the present invention provides a method for optimizing the TFT material manufacturing process. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for optimizing the TFT material manufacturing process to solve the problem that the existing TFT material manufacturing process cannot effectively control the generation of foreign matter and defects during the manufacturing process.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The present invention provides a method for optimizing a TFT material manufacturing process, comprising: Step S101, acquiring TFT material production history data, the TFT material production history data including equipment status history data, process parameter history data, raw material history data, environmental monitoring history data, and foreign matter and defect detection data; Step S102, dividing the TFT material production historical data into a training set and a validation set, using the training set and the validation set to train the random forest model, obtaining a TFT material production equipment control model, receiving a TFT material production task, substituting the TFT material production task into the TFT material production equipment control model, and the TFT material production equipment control model outputs a TFT material production control parameter; Step S103, outputting the TFT material manufacturing control parameters to the TFT material manufacturing equipment, collecting images of the TFT material during the process of the TFT material manufacturing control parameters being executed by the TFT material manufacturing equipment, substituting the collected images into a preset image processing model, detecting foreign matter and defects of the TFT material through the preset image processing model, and obtaining a detection result of foreign matter and defects of the TFT material, wherein the detection result of foreign matter and defects of the TFT material includes a qualified detection of foreign matter and defects of the TFT material and a failed detection of foreign matter and defects of the TFT material; Step S104, matching the TFT material foreign matter and the defect detection failure information in a preset fault defect knowledge base to obtain the device error cause corresponding to the TFT material foreign matter and the defect detection failure information, matching the device error cause corresponding to the TFT material foreign matter and the defect detection failure information in a preset control parameter knowledge base to obtain a control parameter matching result, the control parameter matching result including control parameter adjustment data and matching failure; Step S105, if the control parameter matching result is a failed match, the TFT material production task is substituted into the preset TFT material production task prediction model to generate TFT material production task prediction data, the TFT material production equipment operation data generated in the process of the TFT material production equipment executing the TFT material production control parameters is collected, the TFT material production task prediction data is compared with the TFT material production equipment operation data to obtain the TFT material production equipment operation error data, the TFT material production equipment operation error data and the TFT material production history data are used to optimize the TFT material production equipment control model using a genetic algorithm to obtain an optimized TFT material production equipment control model, the optimized TFT material production task is performed using the optimized TFT material production equipment control model to obtain the optimized TFT material production control parameters, the optimized TFT material production control parameters are sent to the TFT material production equipment, the TFT material generated by the TFT material production equipment executing the optimized TFT material production control parameters is imaged, and the TFT material foreign matter and defects are detected by a preset image processing model, and if the TFT material foreign matter and defect detection is qualified, the optimized TFT material production equipment control model is deployed.

[0006] Furthermore, the method for optimizing the TFT material manufacturing process of the present invention, step S101, comprises: Equipment status history data includes equipment temperature, pressure, humidity, operating hours, and maintenance records; Process parameter history data includes process parameters such as reaction time, temperature setting, pressure setting, and raw material ratio; Raw material historical data includes raw material batch, supplier, purity and composition data; Environmental monitoring historical data includes temperature, humidity and cleanliness environmental parameters of the production workshop; The foreign matter and defect detection data includes the type of foreign matter and defect, the location of foreign matter and defect, the size of foreign matter and defect, and the frequency of occurrence of foreign matter and defect.

[0007] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S102 comprises: From the historical data of TFT material production, a part of the data is randomly selected as the training set for training the random forest model, and the remaining data is used as the validation set; The random forest model is trained using the training set data to obtain the TFT material production equipment control model. The TFT material production equipment control model is used to output TFT material production control parameters according to the input TFT material production tasks. The TFT material production tasks include target output and quality requirements. The TFT material production control parameters include equipment setting parameters and process parameters.

[0008] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S102 comprises: Upon receiving a new TFT material production task, the received TFT material production task data is converted into a model processing format, the new TFT material production task is substituted into the TFT material production equipment control model, and the TFT material production control parameters are output. The TFT material production control parameters include the equipment's working mode, temperature setting, pressure setting, raw material ratio, and reaction time.

[0009] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S103 comprises: According to the TFT material production control parameters, the TFT material production equipment is configured, and after the equipment configuration is completed, the TFT material production process is started, and during the TFT material production process, an image acquisition device is used to acquire images of the TFT material; The TFT material image includes the entire surface and a local area of ​​the TFT material, and the TFT material image data is stored.

[0010] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S103 comprises: The collected TFT material images are preprocessed, and the preprocessed image data is substituted into a preset image processing model. Within the preset image processing model, the TFT material images are analyzed pixel by pixel or region by region to identify abnormal areas that are inconsistent with the normal TFT material surface or structure, and the abnormal areas include defects such as foreign matter in the TFT material, cracks in the TFT material, scratches on the TFT material, and color difference in the TFT material.

[0011] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S104 comprises: Extracting unqualified information from the TFT material foreign matter and defect detection results, the unqualified information includes the type of defect, the location of the defect, the size of the defect and the shape of the defect; Access the preset fault defect knowledge base, which includes various foreign matter and defect types that appear in the TFT material manufacturing process, as well as the correlation between various foreign matter and defect types that appear in the TFT material manufacturing process and the causes of equipment errors; The extracted non-conforming information is compared with the records in the fault defect knowledge base one by one to obtain the fault defect record that best matches the non-conforming information; The corresponding equipment error causes are extracted from the matched fault defect records. The equipment error causes include wear, damage, looseness, inaccurate calibration, and improper control parameter settings of equipment components.

[0012] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S104 comprises: If no item matching the cause of the device error is found in the control parameter knowledge base, the unsuccessful matching result is recorded, and the matching result and control parameter adjustment data are recorded in a document. If the matching is unsuccessful, the control parameter knowledge base is expanded.

[0013] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S105 comprises: The timestamps of the TFT material production task prediction data and the TFT material production equipment operation data are aligned, and each parameter or indicator in the TFT material production task prediction data is compared with the corresponding parameter or indicator in the TFT material production equipment operation data one by one. According to the defined error index, the error data of each comparison is calculated.

[0014] Furthermore, in the method for optimizing the TFT material manufacturing process of the present invention, the step S105 comprises: The optimized TFT material manufacturing control parameters generated by the TFT material manufacturing equipment control model obtained by the genetic algorithm optimization are sent to the TFT material manufacturing equipment. During the process of the TFT material manufacturing equipment executing the optimized control parameters, the generated TFT material is imaged, the image of the acquired TFT material is preprocessed, the preprocessed TFT material image is substituted into a preset image processing model, and the image is analyzed pixel by pixel or region by region using the image processing model to detect foreign matter and defects in the TFT material. Analyze the detection results of the image processing model to determine whether the TFT material is qualified; If the test result is "qualified", that is, no foreign matter or defects are found in the TFT material, proceed to the next step; If the test result is "unqualified", it is necessary to re-analyze the optimized control parameters and optimize or adjust them again; If the foreign matter and defect detection of TFT materials are qualified, the optimized TFT material manufacturing equipment control model is valid; Deploy the optimized control model to the production environment to replace the original control model.

[0015] Beneficial effects of the present invention: Through real-time monitoring and data analysis, the present invention can timely detect and control foreign matter and defects in the TFT material manufacturing process, thereby significantly improving the quality of the TFT material. This helps to reduce the scrap rate, improve the product qualification rate, and further reduce production costs.

[0016] The present invention uses machine learning algorithms (such as random forest models) to learn historical data on TFT material production and generate optimized control parameters. These parameters can more accurately reflect the actual needs of the production process, thereby optimizing the production process. In addition, by continuously optimizing the model through genetic algorithms, it is possible to continuously adapt to changes in production conditions and maintain the optimal state of the production process. Through automated and intelligent production control, the present invention can reduce manual intervention and improve production efficiency. The optimized control parameters can complete production tasks more quickly, shorten the production cycle, and thus improve overall production efficiency.

[0017] The invention helps to reduce the operating costs of enterprises by improving product quality and production efficiency. Reducing the scrap rate and production cycle can save raw materials and energy consumption, while automated and intelligent production control can reduce labor costs.

[0018] The optimization method provided by the present invention can adapt to changes in different production tasks and requirements. By adjusting the control parameters, the production mode can be quickly switched to meet diverse market demands. This flexibility helps companies maintain their competitive advantage in a highly competitive market environment. The implementation and application of the present invention promotes technological innovation in the TFT material manufacturing process. By introducing advanced technologies such as machine learning and image processing, new ideas and methods are provided for the manufacture of TFT materials, which helps to promote technological progress and development of the entire industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0020] Figure 1 A schematic flow chart of a method for optimizing a TFT material manufacturing process provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding 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 creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0022] The present invention provides a method for optimizing a TFT material manufacturing process, comprising: Step S101, acquiring TFT material production history data, the TFT material production history data including equipment status history data, process parameter history data, raw material history data, environmental monitoring history data, and foreign matter and defect detection data; Equipment status historical data includes equipment temperature, pressure, humidity, operating hours, and maintenance records, which reflect the working status and health of the equipment.

[0023] The historical data of process parameters include reaction time, temperature setting, pressure setting and raw material ratio, etc. These data directly affect the production quality of TFT materials.

[0024] The historical data of raw materials include the batch, supplier, purity and composition data of the raw materials. The quality of the raw materials is also an important factor affecting the performance of TFT materials.

[0025] The historical environmental monitoring data includes the temperature, humidity and cleanliness parameters of the production workshop. Environmental factors also have an important impact on the production of TFT materials.

[0026] Foreign matter and defect detection data includes the type, location, size and frequency of occurrence of foreign matter and defects, which are used for subsequent analysis and optimization.

[0027] Step S102, dividing the TFT material production historical data into a training set and a validation set, using the training set and the validation set to train the random forest model, obtaining a TFT material production equipment control model, receiving a TFT material production task, substituting the TFT material production task into the TFT material production equipment control model, and the TFT material production equipment control model outputs a TFT material production control parameter; The acquired historical data is divided into a training set and a validation set.

[0028] The random forest model is trained using the training set data to obtain the TFT material production equipment control model. The model can output the corresponding TFT material production control parameters (such as equipment setting parameters and process parameters) based on the input TFT material production tasks (including target output and quality requirements).

[0029] When a new TFT material production task is received, it is converted into a model processing format, substituted into the trained model, and the control parameters are output.

[0030] Step S103, outputting the TFT material manufacturing control parameters to the TFT material manufacturing equipment, collecting images of the TFT material during the process of the TFT material manufacturing control parameters being executed by the TFT material manufacturing equipment, substituting the collected images into a preset image processing model, detecting foreign matter and defects of the TFT material through the preset image processing model, and obtaining a detection result of foreign matter and defects of the TFT material, wherein the detection result of foreign matter and defects of the TFT material includes a qualified detection of foreign matter and defects of the TFT material and a failed detection of foreign matter and defects of the TFT material; The output control parameters are sent to the TFT material production equipment, and image acquisition is performed during the equipment execution process.

[0031] The collected TFT material images include the overall surface and local areas for subsequent analysis.

[0032] The collected images are preprocessed and substituted into the preset image processing model. The model identifies abnormal areas that do not conform to normal TFT materials, such as foreign matter, cracks, scratches, and color differences, through pixel-by-pixel or region-by-region analysis.

[0033] Based on the output results of the image processing model, it is determined whether the TFT material foreign matter and defect detection is qualified.

[0034] Step S104, matching the TFT material foreign matter and the defect detection failure information in a preset fault defect knowledge base to obtain the device error cause corresponding to the TFT material foreign matter and the defect detection failure information, matching the device error cause corresponding to the TFT material foreign matter and the defect detection failure information in a preset control parameter knowledge base to obtain a control parameter matching result, the control parameter matching result including control parameter adjustment data and matching failure; For the information of unqualified detection, key information such as defect type, location, size and shape is extracted.

[0035] Match the faults in the preset knowledge base to find the corresponding causes of equipment errors, which may include wear, damage, looseness, inaccurate calibration or improper setting of control parameters.

[0036] The matched device error causes are further matched in a preset control parameter knowledge base to try to find the corresponding control parameter adjustment data.

[0037] If the match is successful, the control parameters are optimized according to the adjustment data; if the match is unsuccessful, proceed to the next step.

[0038] Step S105, if the control parameter matching result is a failed match, the TFT material production task is substituted into the preset TFT material production task prediction model to generate TFT material production task prediction data, the TFT material production equipment operation data generated in the process of the TFT material production equipment executing the TFT material production control parameters is collected, the TFT material production task prediction data is compared with the TFT material production equipment operation data to obtain the TFT material production equipment operation error data, the TFT material production equipment operation error data and the TFT material production history data are used to optimize the TFT material production equipment control model using a genetic algorithm to obtain an optimized TFT material production equipment control model, the optimized TFT material production task is performed using the optimized TFT material production equipment control model to obtain the optimized TFT material production control parameters, the optimized TFT material production control parameters are sent to the TFT material production equipment, the TFT material generated by the TFT material production equipment executing the optimized TFT material production control parameters is imaged, and the TFT material foreign matter and defects are detected by a preset image processing model, and if the TFT material foreign matter and defect detection is qualified, the optimized TFT material production equipment control model is deployed.

[0039] In the case of unsuccessful matching, the TFT material production task is substituted into the preset TFT material production task prediction model to generate prediction data.

[0040] At the same time, the operating data of the TFT material production equipment during the execution of control parameters is collected.

[0041] The predicted data are compared with the operating data, the error data are calculated, and the genetic algorithm is used to optimize the control model of the TFT material manufacturing equipment.

[0042] The optimized model is used again to generate control parameters and is verified on the TFT material manufacturing equipment. If the verification result is qualified, the optimized model is deployed to the production environment to replace the original control model.

[0043] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S101, comprises: Equipment status history data includes equipment temperature, pressure, humidity, operating hours, and maintenance records; Process parameter history data includes process parameters such as reaction time, temperature setting, pressure setting, and raw material ratio; Raw material historical data includes raw material batch, supplier, purity and composition data; Environmental monitoring historical data includes temperature, humidity and cleanliness environmental parameters of the production workshop; The foreign matter and defect detection data includes the type of foreign matter and defect, the location of foreign matter and defect, the size of foreign matter and defect, and the frequency of occurrence of foreign matter and defect.

[0044] Equipment status history data: The temperature records the actual operating temperature of the equipment at different operation stages (such as preheating, processing, cooling, etc.), as well as the temperature fluctuation range, which is used to evaluate the impact of equipment thermal stability on material properties.

[0045] Pressure monitoring measures the pressure changes in the reaction chamber or processing cavity to ensure that the pressure is maintained within the set range to control the gas atmosphere and reaction rate.

[0046] Humidity Especially when it comes to moisture-sensitive materials, accurately record the humidity level of the production environment to avoid adverse effects of moisture on the materials.

[0047] Running time accumulates the running time of each component of the equipment, combines it with the maintenance cycle, predicts potential failure points, and implements preventive maintenance.

[0048] Maintenance records the content, time, replaced parts and reasons of each maintenance, providing a basis for equipment performance optimization and troubleshooting.

[0049] Process parameter historical data: Reaction time: Control the time of each step of chemical reaction or physical treatment to make the material structure reach the expected level.

[0050] Temperature setting: According to material characteristics and reaction requirements, set and strictly monitor the temperature curve during processing.

[0051] Pressure Setting: Adjusting and maintaining specific pressure conditions to promote or inhibit specific chemical reactions or physical changes.

[0052] Raw material ratio: Based on experimental data and production experience, optimize the mixing ratio of raw materials to ensure product consistency and high performance.

[0053] Raw material historical data: Batch: Track the origin and use of each batch of raw materials to facilitate tracing the cause when problems arise.

[0054] Suppliers: Record raw material supplier information, evaluate the stability of raw material quality from different suppliers, and switch or optimize suppliers when necessary.

[0055] Purity: Regularly test the purity of raw materials to ensure that impurities are within acceptable ranges to avoid negative impact on product quality.

[0056] Composition data: Detailed analysis of the chemical composition of raw materials, especially the content and proportion of key ingredients, to meet production requirements.

[0057] Environmental monitoring historical data: Temperature and humidity in the production workshop: Maintain constant temperature and humidity conditions in the production environment to reduce the interference of external environmental changes on the production process.

[0058] Cleanliness: Regularly check the air cleanliness in the production area to prevent dust particles from contaminating the product, which is especially important for high-precision TFT materials.

[0059] Foreign matter and defect detection data: Type: Categorize and record the types of detected foreign matter (such as dust, metal particles) and defects (such as cracks, color difference) to facilitate cause analysis.

[0060] Location: Accurately mark the specific location of foreign matter or defects on the product to provide guidance for subsequent repairs or quality control.

[0061] Size: Measure and record the size of foreign matter and defects to assess how they may affect product performance.

[0062] Occurrence frequency: Count the occurrence frequency of various foreign matter and defects, analyze the trend, and adjust the production process or strengthen quality control measures in time to reduce the defective rate.

[0063] Through the above detailed data collection and analysis, the present invention can comprehensively monitor various variables in the TFT material manufacturing process, timely discover and solve problems, thereby effectively improving product quality and production efficiency.

[0064] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S102 includes: From the historical data of TFT material production, a part of the data is randomly selected as the training set for training the random forest model, and the remaining data is used as the validation set; The random forest model is trained using the training set data to obtain the TFT material production equipment control model. The TFT material production equipment control model is used to output TFT material production control parameters according to the input TFT material production tasks. The TFT material production tasks include target output and quality requirements. The TFT material production control parameters include equipment setting parameters and process parameters.

[0065] From the historical data of TFT material production, a part of the data is randomly selected as the training set, taking into account the time span, representativeness, and completeness of the data. This part of the data should cover a variety of production conditions, equipment status, raw material batches, etc. to ensure the generalization ability of the model.

[0066] Clean the selected training set data, remove outliers, missing values, etc., and perform necessary normalization or standardization to eliminate dimensional differences between data and improve model training efficiency.

[0067] The cleaned data is divided into a certain proportion (e.g. 70% training set, 30% validation set). The training set is used to train the random forest model, and the validation set is used to evaluate the performance and generalization ability of the model.

[0068] Based on the training set data, a random forest model is constructed. Random forest is an integrated learning method that improves the accuracy and stability of the model by constructing multiple decision trees and integrating their prediction results. Key features such as equipment state parameters (temperature, pressure, etc.), process parameters (reaction time, raw material ratio, etc.), raw material characteristics (batch, purity, etc.) and environmental parameters (temperature, humidity, etc.) are extracted from the historical data of TFT material production as input variables of the model. Through methods such as cross-validation, the parameters of the random forest model (such as the number of trees, maximum depth, etc.) are adjusted to find the model configuration with the best performance.

[0069] The validation set data is used to evaluate the performance of the trained random forest model, including indicators such as accuracy, recall, F1 score, and the model's predictive ability for TFT material production tasks (such as target output and quality requirements).

[0070] According to the performance evaluation results, the model is adjusted and optimized as necessary, for example, adding or deleting features, adjusting model parameters, adopting a more complex model structure, etc., to improve the prediction accuracy and generalization ability of the model.

[0071] According to production requirements, the target output and quality requirements of TFT material production are input. These requirements will be used as input information for the model to generate corresponding control parameters.

[0072] Output TFT material production control parameters. Based on the input task requirements, the model combines historical data and machine learning algorithms to output TFT material production control parameters, including equipment setting parameters (such as temperature setting value, pressure setting value, etc.) and process parameters (such as reaction time, raw material ratio, etc.).

[0073] The control parameters output by the model are applied to TFT material production equipment to achieve automated and precise production control, thereby improving production efficiency and product quality.

[0074] Through the above steps, the present invention uses the random forest algorithm to establish a TFT material manufacturing equipment control model, achieves the goal of automatically outputting optimal control parameters according to production tasks, and provides strong support for the efficient and high-quality production of TFT materials.

[0075] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S102 includes: Upon receiving a new TFT material production task, the received TFT material production task data is converted into a model processing format, the new TFT material production task is substituted into the TFT material production equipment control model, and the TFT material production control parameters are output. The TFT material production control parameters include the equipment's working mode, temperature setting, pressure setting, raw material ratio, and reaction time.

[0076] Receive new TFT material production tasks: New TFT material production tasks may come from the production planning system, customer orders or R&D department's experimental requirements. The task should clearly specify the type, specification, target output, quality requirements and other key information of the TFT material.

[0077] Convert the received TFT material production task data into a format that the model can process. This may involve data type conversion (such as text to numbers), data structure adjustment (such as expanding multidimensional data into a one-dimensional array), etc. Extract features related to model input from the task data, such as target output, quality requirements, etc., and ensure that these features are consistent with the features used in model training. Verify the converted data to ensure data integrity and accuracy, and avoid inaccurate control parameters of the model output due to data errors.

[0078] Substitute into the TFT material production equipment control model: Model loading: First load the trained TFT material production equipment control model to ensure that the model is in a usable state.

[0079] Data input: The pre-processed TFT material production task data is used as the input of the model and substituted into the model.

[0080] Model reasoning: The model uses the learned knowledge and rules to perform reasoning calculations based on the input data and generate control parameters for TFT material production.

[0081] Control parameter content: The control parameters output by the model should include the equipment's operating mode (such as continuous production, batch production, etc.), temperature settings (such as preheating temperature, processing temperature, etc.), pressure settings (such as reaction chamber pressure, delivery pressure, etc.), raw material ratio (such as the feed ratio of each raw material) and reaction time (such as total reaction time, reaction time of each stage, etc.).

[0082] Parameter verification: Verify the output control parameters to ensure that they are feasible and safe in actual production. For example, check whether the temperature setting is within the safe working range of the equipment, whether the raw material ratio meets the requirements of the chemical reaction, etc.

[0083] Parameter transfer: The verified control parameters are transferred to the production control system or operators to guide the actual production of TFT materials.

[0084] Record the new TFT material production tasks and their corresponding control parameters, production results and other information for subsequent data analysis and model optimization. Establish a feedback mechanism to collect actual data from the production process (such as equipment status, product quality, etc.) to evaluate the performance and accuracy of the model and provide data support for the continuous improvement of the model.

[0085] Through the above steps, the present invention realizes the conversion of new TFT material production tasks into specific control parameters and applies them to the actual production process, thereby improving the automation level and production efficiency of TFT material production.

[0086] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S103 includes: According to the TFT material production control parameters, the TFT material production equipment is configured, and after the equipment configuration is completed, the TFT material production process is started, and during the TFT material production process, an image acquisition device is used to acquire images of the TFT material; The TFT material image includes the entire surface and a local area of ​​the TFT material, and the TFT material image data is stored.

[0087] The collected TFT material images are preprocessed, and the preprocessed image data is substituted into a preset image processing model. Within the preset image processing model, the TFT material images are analyzed pixel by pixel or region by region to identify abnormal areas that are inconsistent with the normal TFT material surface or structure, and the abnormal areas include defects such as foreign matter in the TFT material, cracks in the TFT material, scratches on the TFT material, and color difference in the TFT material.

[0088] According to the TFT material production control parameters output in step S102, the TFT material production equipment is precisely configured, including setting the working mode of the equipment, adjusting the temperature and pressure to predetermined values, preparing raw materials of a specific ratio, and setting the reaction time.

[0089] Production start: After the equipment configuration is completed, the production process of TFT materials is started. This process may include multiple steps such as mixing, reaction, deposition, annealing, etc. of raw materials, depending on the type of TFT material and the production process.

[0090] Image acquisition equipment selection: Select high-resolution, high-sensitivity image acquisition equipment, such as industrial cameras or microscopes, to ensure that the surface and structural details of the TFT material can be clearly captured.

[0091] Acquisition strategy: Develop an image acquisition strategy, including acquisition frequency, acquisition location (whole surface and local area), lighting conditions, etc., to ensure that the acquired images can fully reflect the condition of the TFT material.

[0092] Image storage: Store the acquired TFT material images in a suitable format (such as JPEG, PNG, etc.) in a designated database or file system for subsequent processing and analysis.

[0093] Image preprocessing: Apply image denoising algorithms, such as Gaussian filtering and mean filtering, to remove noise and interference in the image and improve image quality. Perform image enhancement processing, such as contrast enhancement and sharpening, to make the surface structure and defects of TFT materials more visible. If a local area needs to be analyzed separately, an image segmentation algorithm can be applied to separate the area of ​​interest from the background.

[0094] Select or develop image processing models suitable for TFT material defect identification, such as convolutional neural networks (CNNs), support vector machines (SVMs), etc. These models can automatically learn and identify various defects on the surface of TFT materials. Use the preprocessed image data as the input of the model and substitute it into the preset image processing model. Pixel-by-pixel / region-by-region analysis: Within the model, the TFT material image is analyzed pixel-by-pixel or region-by-region. This includes extracting image features, calculating feature values, and comparing with the features of normal TFT material surfaces or structures.

[0095] Through the analysis and comparison of the model, abnormal areas that do not conform to the normal TFT material surface or structure are identified. These abnormal areas may include defects such as foreign matter, cracks, scratches, and color difference. The identified defects are classified (such as foreign matter, cracks, etc.) and marked (such as marking the location and size of the defect on the image). This helps to count, analyze and process the defects later.

[0096] The results of defect identification are displayed in an intuitive way, such as generating defect distribution diagrams, defect statistical reports, etc. A feedback mechanism is established to feed back the results of defect identification to the production control system or operators so that the production process can be adjusted in time, equipment failures can be repaired, or other corrective measures can be taken. At the same time, the results can also be used for continuous improvement of the model and update of training data.

[0097] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S104 includes: Extracting unqualified information from the TFT material foreign matter and defect detection results, the unqualified information includes the type of defect, the location of the defect, the size of the defect and the shape of the defect; Access the preset fault defect knowledge base, which includes various foreign matter and defect types that appear in the TFT material manufacturing process, as well as the correlation between various foreign matter and defect types that appear in the TFT material manufacturing process and the causes of equipment errors; The extracted non-conforming information is compared with the records in the fault defect knowledge base one by one to obtain the fault defect record that best matches the non-conforming information; The corresponding equipment error causes are extracted from the matched fault defect records. The equipment error causes include wear, damage, looseness, inaccurate calibration, and improper control parameter settings of equipment components.

[0098] If no item matching the cause of the device error is found in the control parameter knowledge base, the unsuccessful matching result is recorded, and the matching result and control parameter adjustment data are recorded in a document. If the matching is unsuccessful, the control parameter knowledge base is expanded.

[0099] Extract unqualified information from the TFT material foreign matter and defect detection results. This information is usually obtained through image processing models or manual inspection. Unqualified information should include the type of defect (such as foreign matter, cracks, scratches, color difference, etc.), the location of the defect (specific coordinates or area on the TFT material), the size of the defect (such as length, width, area, etc.) and the morphology of the defect (such as shape, color, etc.).

[0100] The preset fault defect knowledge base should include various foreign matter and defect types that may appear in the TFT material production process, as well as the relationship between these foreign matter and defect types and the cause of equipment errors. The knowledge base can be stored and managed in a database, expert system or other forms. Access the fault defect knowledge base through a programming interface or query statement to obtain data related to non-conforming information.

[0101] Matching nonconformance information with fault defect records: Matching algorithm: Use algorithms such as similarity calculation, pattern matching or rule reasoning to compare the extracted non-conforming information with the records in the fault defect knowledge base one by one to find the fault defect record that best matches the non-conforming information.

[0102] Matching criteria: Set matching thresholds or rules to determine when two records are considered a match. For example, they can be considered a match when the defect type, location, size, and morphology are similar or identical.

[0103] Extract device error reason: Cause identification: Extract the corresponding equipment error cause from the matched fault defect records. The equipment error cause may include wear, damage, looseness, inaccurate calibration, improper control parameter settings, etc.

[0104] Cause classification: Classify and encode the extracted device error causes to facilitate subsequent processing and analysis.

[0105] Handling unsuccessful matching: Check the control parameter knowledge base: First, check whether there is a control parameter knowledge base that contains the correspondence between the causes of device errors and the adjustable control parameters.

[0106] Match attempt: An attempt is made to find a match to the cause of the device error in the control parameter knowledge base. If a match is found, the control parameters can be adjusted based on this knowledge and recorded in the document.

[0107] Record unmatched results: If no item matching the cause of the device error is found in the control parameter knowledge base, the unsuccessful matching result is recorded, including the cause of the mismatch, the current control parameter settings and other information.

[0108] Expanding the knowledge base: For unsuccessful matching situations, new device error causes, corresponding control parameter adjustment suggestions, and related TFT material foreign matter and defect detection results can be added to the fault defect knowledge base and control parameter knowledge base to enrich and expand the content of the knowledge base.

[0109] The matching results, control parameter adjustment data, and any mismatches and subsequent treatment measures are recorded in documents for subsequent review and tracking. A feedback mechanism is established to feed back the matching results and control parameter adjustment suggestions to the production control system or operators so that they can adjust the production process, repair equipment failures, or take other corrective measures in a timely manner.

[0110] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S105, includes: The timestamps of the TFT material production task prediction data and the TFT material production equipment operation data are aligned, and each parameter or indicator in the TFT material production task prediction data is compared with the corresponding parameter or indicator in the TFT material production equipment operation data one by one. According to the defined error index, the error data of each comparison is calculated.

[0111] Obtain prediction data from the TFT material production task prediction system. These data usually include predicted production volume, production efficiency, energy consumption, equipment status and other parameters or indicators, as well as their changing trends over time. Obtain actual operation data from the monitoring system of the TFT material production equipment. These data should cover the actual production volume, production efficiency, energy consumption, equipment status and other parameters or indicators corresponding to the predicted data, and contain timestamp information to record the data collection time.

[0112] Timestamp alignment: Time synchronization: The timestamps of the forecast data and the running data are synchronized. If there is a time difference between the two, time correction is required to ensure the accuracy of the comparison.

[0113] Data alignment: Align the forecast data with the running data based on the timestamp information. For each time point, ensure that there are corresponding forecast values ​​and actual values ​​for comparison.

[0114] Parameter or indicator comparison: Item-by-item comparison: Compare each parameter or indicator in the forecast data with the corresponding parameter or indicator in the operating data one by one. This includes the forecast value and actual value of production volume, the forecast value and actual value of production efficiency, the forecast value and actual value of energy consumption, etc.

[0115] Data preprocessing: Before comparison, you may need to preprocess the data, such as removing outliers, smoothing data, etc., to ensure the fairness and accuracy of the comparison.

[0116] Error calculation: Define error indicators: Define the indicators for error calculation according to actual needs. Common error indicators include absolute error, relative error, mean square error, etc.

[0117] Calculate the error: Calculate the error data of each comparison according to the defined error index. For example, for the predicted value and actual value of production volume, their absolute error or relative error can be calculated; for the predicted value and actual value of production efficiency, their mean square error can be calculated.

[0118] Error analysis: Analyze the calculated error data to identify the parameters or indicators with large errors and their possible causes. This helps to optimize and adjust the prediction model or equipment operating status in the future.

[0119] Result output and feedback: Result display: Display the comparison results and error data in an intuitive way, such as generating comparison charts, error distribution charts, etc.

[0120] Feedback mechanism: Establish a feedback mechanism to feed back comparison results and error data to relevant personnel or systems so that they can promptly understand the prediction accuracy of TFT material production tasks and equipment operating status, and make adjustments and optimizations as needed.

[0121] Through the above steps, the present invention realizes the detailed comparison and error calculation between the prediction data of the TFT material production task and the equipment operation data, and provides strong data support for evaluating the accuracy of the prediction model and optimizing the equipment operation status.

[0122] Specifically, the method for optimizing the TFT material manufacturing process of the present invention, step S105, includes: The optimized TFT material manufacturing control parameters generated by the TFT material manufacturing equipment control model obtained by the genetic algorithm optimization are sent to the TFT material manufacturing equipment. During the process of the TFT material manufacturing equipment executing the optimized control parameters, the generated TFT material is imaged, the image of the acquired TFT material is preprocessed, the preprocessed TFT material image is substituted into a preset image processing model, and the image is analyzed pixel by pixel or region by region using the image processing model to detect foreign matter and defects in the TFT material. Analyze the detection results of the image processing model to determine whether the TFT material is qualified; If the test result is "qualified", that is, no foreign matter or defects are found in the TFT material, proceed to the next step; If the test result is "unqualified", it is necessary to re-analyze the optimized control parameters and optimize or adjust them again; If the foreign matter and defect detection of TFT materials are qualified, the optimized TFT material manufacturing equipment control model is valid; Deploy the optimized control model to the production environment to replace the original control model.

[0123] Specifically, the method for optimizing the TFT (thin film transistor) material manufacturing process described in the present invention, in step S105, describes in detail how to apply the optimized control parameters to the TFT material manufacturing equipment, and detect foreign matter and defects in the generated TFT material through an image processing model, and perform subsequent processing based on the detection results.

[0124] The equipment operation error data and historical data are input into the genetic algorithm, the parameter combination is evaluated through the fitness function, and the optimization model is iteratively generated.

[0125] Generation of optimized control parameters: First, the TFT material manufacturing equipment control model is optimized through genetic algorithms to obtain optimized TFT material manufacturing control parameters. These parameters may include temperature, pressure, speed, time, etc., depending on the process requirements of TFT material manufacturing.

[0126] Parameter transmission: The optimized control parameters are sent to the control system of the TFT material production equipment. This is usually achieved through a communication interface or network to ensure that the control parameters can be accurately and timely transmitted to the equipment.

[0127] Parameter application: The TFT material production equipment is adjusted according to the received optimized control parameters, and the TFT material is produced according to the new parameter settings.

[0128] TFT material image acquisition and preprocessing: Image acquisition: During or after the TFT material production process, the produced TFT material is imaged. This is usually done using a high-resolution camera or imaging system to ensure that the subtle features on the surface of the TFT material can be captured.

[0129] Image preprocessing: Preprocess the collected TFT material images, including denoising, enhancement, correction and other steps. The purpose of preprocessing is to improve the quality of the image and make the subsequent image analysis more accurate.

[0130] Image processing and foreign body and defect detection: Image processing model: Substitute the pre-processed TFT material image into a preset image processing model. The image processing model may be built based on machine learning, deep learning or traditional image processing algorithms to analyze the image pixel by pixel or region by region.

[0131] Foreign matter and defect detection: Use image processing models to perform detailed analysis on images to detect foreign matter (such as dust, particles) and defects (such as cracks, scratches, color difference, etc.) in TFT materials. The detection process may include steps such as feature extraction, classification, and recognition.

[0132] Analysis and processing of test results: Qualification judgment: According to the detection results of the image processing model, judge whether the TFT material is qualified. If the test result is "qualified", that is, no foreign matter and defects are found in the TFT material, then continue to the next process flow.

[0133] Failure treatment: If the test result is "failed", the optimized control parameters need to be re-analyzed. This may include reviewing the optimization process of the genetic algorithm, checking the setting range of the control parameters, considering other factors that may affect the quality of the TFT material, etc. Based on the analysis results, the control parameters are optimized or adjusted again and re-applied to the TFT material manufacturing equipment.

[0134] Validation and deployment of control models: Validity verification: If after multiple optimizations and adjustments, the foreign matter and defect detection of TFT materials are qualified, then the optimized TFT material manufacturing equipment control model can be considered effective. At this time, the control model needs to be fully verified and tested to ensure its stability and reliability in actual production.

[0135] Model deployment: Deploy the optimized control model to the production environment to replace the original control model. This usually involves steps such as integrating the new control model into the production control system, updating relevant configuration files, and training operators. Ensure that the new control model can run smoothly in the production environment and continuously monitor its performance to ensure high-quality production of TFT materials.

[0136] The technical solution of the present invention solves the problem that the existing TFT material manufacturing process cannot effectively control the generation of foreign matter and defects during the manufacturing process through the following steps: Obtain historical data on TFT material production, including equipment status, process parameters, raw material information, environmental monitoring data, and foreign matter and defect detection data. Use the data to train a random forest model to obtain a TFT material production equipment control model. The model can output corresponding control parameters based on new TFT material production tasks.

[0137] During the TFT material production process, images of TFT materials are collected in real time. The collected images are substituted into the preset image processing model, and foreign matter and defects in the TFT material are detected through model analysis. If foreign matter and defects are detected, unqualified information is extracted and matched in the fault defect knowledge base to find the corresponding equipment error cause.

[0138] Match in the control parameter knowledge base and try to find control parameter adjustment data to optimize the production process. If no match is found in the knowledge base, substitute the TFT material production task into the prediction model to generate prediction data, compare it with the equipment operation data, and calculate the error. Use genetic algorithms to optimize the control model based on the error data and generate new control parameters. Apply the optimized control parameters to actual production, and perform image acquisition and defect detection again to verify the optimization effect. If the verification is qualified, deploy the optimized control model to the production environment and replace the original control model. By continuously collecting new production data, the model is continuously trained and optimized to continuously improve the quality and production efficiency of TFT materials.

[0139] Through the above technical solution, the present invention can monitor and control the generation of foreign matter and defects in real time during the TFT material production process, effectively improving product quality and production efficiency.

Claims

1. A method for optimizing a TFT material manufacturing process, characterized in that: include: Step S101, acquiring TFT material production history data, the TFT material production history data including equipment status history data, process parameter history data, raw material history data, environmental monitoring history data, and foreign matter and defect detection data; Step S102, dividing the TFT material production historical data into a training set and a validation set, using the training set and the validation set to train the random forest model, obtaining a TFT material production equipment control model, receiving a TFT material production task, substituting the TFT material production task into the TFT material production equipment control model, and the TFT material production equipment control model outputs a TFT material production control parameter; Step S103, outputting the TFT material manufacturing control parameters to the TFT material manufacturing equipment, collecting images of the TFT material during the process of the TFT material manufacturing control parameters being executed by the TFT material manufacturing equipment, substituting the collected images into a preset image processing model, detecting foreign matter and defects of the TFT material through the preset image processing model, and obtaining a detection result of foreign matter and defects of the TFT material, wherein the detection result of foreign matter and defects of the TFT material includes a qualified detection of foreign matter and defects of the TFT material and a failed detection of foreign matter and defects of the TFT material; Step S104, matching the TFT material foreign matter and the defect detection failure information in a preset fault defect knowledge base, obtaining the equipment error cause corresponding to the TFT material foreign matter and the defect detection failure information, matching the equipment error cause corresponding to the TFT material foreign matter and the defect detection failure information in a preset control parameter knowledge base, obtaining a control parameter matching result, the control parameter matching result including control parameter adjustment data and matching failure, the fault defect knowledge base including a correlation database of historical defect types, equipment error causes and solutions, constructed through expert experience and production data; Step S105, if the control parameter matching result is a failed match, the TFT material production task is substituted into the preset TFT material production task prediction model to generate TFT material production task prediction data, the TFT material production equipment operation data generated in the process of the TFT material production equipment executing the TFT material production control parameters is collected, the TFT material production task prediction data is compared with the TFT material production equipment operation data to obtain the TFT material production equipment operation error data, the TFT material production equipment operation error data and the TFT material production history data are used to optimize the TFT material production equipment control model using a genetic algorithm to obtain an optimized TFT material production equipment control model, the optimized TFT material production task is performed using the optimized TFT material production equipment control model to obtain the optimized TFT material production control parameters, the optimized TFT material production control parameters are sent to the TFT material production equipment, the TFT material generated by the TFT material production equipment executing the optimized TFT material production control parameters is imaged, and the TFT material foreign matter and defects are detected by a preset image processing model, and if the TFT material foreign matter and defect detection is qualified, the optimized TFT material production equipment control model is deployed.

2. The method for optimizing the TFT material manufacturing process according to claim 1, characterized in that: The step S101 includes: Equipment status history data includes equipment temperature, pressure, humidity, operating hours, and maintenance records; Process parameter history data includes process parameters such as reaction time, temperature setting, pressure setting, and raw material ratio; Raw material historical data includes raw material batch, supplier, purity and composition data; Environmental monitoring historical data includes temperature, humidity and cleanliness environmental parameters of the production workshop; The foreign matter and defect detection data includes the type of foreign matter and defect, the location of foreign matter and defect, the size of foreign matter and defect, and the frequency of occurrence of foreign matter and defect.

3. The method for optimizing the TFT material manufacturing process according to claim 1, characterized in that: The step S102 includes: From the historical data of TFT material production, a part of the data is randomly selected as the training set for training the random forest model, and the remaining data is used as the validation set; The random forest model is trained using the training set data to obtain the TFT material manufacturing equipment control model. The TFT material production equipment control model is used to output TFT material production control parameters according to the input TFT material production tasks. The TFT material production tasks include target output and quality requirements. The TFT material production control parameters include equipment setting parameters and process parameters.

4. The method for optimizing the TFT material manufacturing process according to claim 3, characterized in that: The step S102 includes: Upon receiving a new TFT material production task, the received TFT material production task data is converted into a model processing format, the new TFT material production task is substituted into the TFT material production equipment control model, and the TFT material production control parameters are output. The TFT material production control parameters include the equipment's working mode, temperature setting, pressure setting, raw material ratio, and reaction time.

5. The method for optimizing TFT material manufacturing process according to claim 1, characterized in that: The step S103 includes: According to the TFT material production control parameters, the TFT material production equipment is configured, and after the equipment configuration is completed, the TFT material production process is started, and during the TFT material production process, an image acquisition device is used to acquire images of the TFT material; The TFT material image includes the entire surface and a local area of ​​the TFT material, and the TFT material image data is stored.

6. The method for optimizing TFT material manufacturing process according to claim 5, characterized in that: The step S103 includes: The collected TFT material images are preprocessed, and the preprocessed image data is substituted into a preset image processing model. Within the preset image processing model, the TFT material images are analyzed pixel by pixel or region by region to identify abnormal areas that are inconsistent with the normal TFT material surface or structure, and the abnormal areas include defects such as foreign matter in the TFT material, cracks in the TFT material, scratches on the TFT material, and color difference in the TFT material.

7. The method for optimizing TFT material manufacturing process according to claim 1, characterized in that: The step S104 includes: Extracting unqualified information from the TFT material foreign matter and defect detection results, the unqualified information includes the type of defect, the location of the defect, the size of the defect and the shape of the defect; Access the preset fault defect knowledge base, which includes various foreign matter and defect types that appear in the TFT material manufacturing process, as well as the correlation between various foreign matter and defect types that appear in the TFT material manufacturing process and the causes of equipment errors; The extracted non-conforming information is compared with the records in the fault defect knowledge base one by one to obtain the fault defect record that best matches the non-conforming information; The corresponding equipment error causes are extracted from the matched fault defect records. The equipment error causes include wear, damage, looseness, inaccurate calibration, and improper control parameter settings of equipment components.

8. The method for optimizing TFT material manufacturing process according to claim 7, characterized in that: The step S104 includes: If no item matching the cause of the device error is found in the control parameter knowledge base, the unsuccessful matching result is recorded, and the matching result and control parameter adjustment data are recorded in a document. If the matching is unsuccessful, the control parameter knowledge base is expanded.

9. The method for optimizing TFT material manufacturing process according to claim 1, characterized in that: The step S105 includes: The timestamps of the TFT material production task prediction data and the TFT material production equipment operation data are aligned, and each parameter or indicator in the TFT material production task prediction data is compared with the corresponding parameter or indicator in the TFT material production equipment operation data one by one. According to the defined error index, the error data of each comparison is calculated.

10. The method for optimizing TFT material manufacturing process according to claim 9, characterized in that: The step S105 includes: The optimized TFT material manufacturing control parameters generated by the TFT material manufacturing equipment control model obtained by the genetic algorithm optimization are sent to the TFT material manufacturing equipment. During the process of the TFT material manufacturing equipment executing the optimized control parameters, the generated TFT material is imaged, the image of the acquired TFT material is preprocessed, the preprocessed TFT material image is substituted into a preset image processing model, and the image is analyzed pixel by pixel or region by region using the image processing model to detect foreign matter and defects in the TFT material. Analyze the detection results of the image processing model to determine whether the TFT material is qualified; If the test result is "qualified", that is, no foreign matter or defects are found in the TFT material, proceed to the next step; If the test result is "unqualified", it is necessary to re-analyze the optimized control parameters and optimize or adjust them again; If the foreign matter and defect detection of TFT materials are qualified, the optimized TFT material manufacturing equipment control model is valid; Deploy the optimized control model to the production environment to replace the original control model.

Images