Intelligent control laser transverse wiredrawing transfer printing method and system
Through intelligent control methods and deep learning models, the laser transfer printing process parameters are optimized, and the product inconsistency and quality instability caused by process parameters are solved, and the production efficiency and product quality are improved.
Patent Information
- Application Number
- CN202410948275.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-08
AI Technical Summary
The process parameters in the laser transfer printing process are complex and highly coupled, which makes manual parameters time-consuming and labor-consuming, prone to artificial errors, and it is difficult to ensure product consistency and quality stability.
Through intelligent control methods, the product type and preset indicator performance parameters of the product to be printed are determined, and the initial process parameters are retrieved from the laser printing process database in combination with the performance testing steps, laser printing processing and performance testing are carried out, and the two-way long and short-term memory network and diagrams are used. Neural networks build parameter correlation analysis models and performance prediction models, and optimize process parameters to achieve optimality.
Systematized and data-driven process parameter optimization is achieved, reducing manual adjustment time and error, improving the consistency and stability of production efficiency and product quality, and reducing production costs and resource waste.
Smart Images

Figure CN118917649A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of laser printing, and in particular relates to an intelligently controlled laser transverse wire drawing transfer printing method and system. Background Art
[0002] Laser lateral drawing transfer printing technology is a high-precision manufacturing technology that uses laser energy to transfer materials from one substrate to another. This technology combines the advantages of laser processing and printing technology, and can achieve high-resolution, non-contact material transfer and patterning at the micron or even nanometer scale. The fully automatic laser lateral drawing transfer printing process production line is a highly integrated manufacturing system that combines technologies such as automated control, precision machinery, laser processing, image processing and data management to achieve efficient and high-precision material transfer and patterning operations. The core of the production line is the automated control system. By inputting process parameters into the automated control system, the process can be fully automated.
[0003] Due to the different sensitivity of different materials and product types to process parameters, in order to ensure the quality of each type of product, it is usually necessary to manually adjust the parameters for each product independently. However, the laser transfer printing process often involves multiple complex process parameters, which are highly coupled and nonlinear, making it difficult to accurately optimize them through simple empirical formulas or single tests. As a result, parameter adjustment is not only time-consuming and labor-intensive, increasing production costs, but also prone to human errors and instability when manually adjusting parameters, making it difficult to ensure product consistency and quality stability. Summary of the invention
[0004] The present invention provides an intelligently controlled laser transverse wire drawing transfer printing method and system to solve the problem that manual adjustment of parameters is prone to human errors and instability, making it difficult to ensure product consistency and quality stability.
[0005] In a first aspect, the present invention provides an intelligently controlled laser lateral wire drawing transfer printing method, the method comprising the following steps: Determine the product type and preset performance parameters of the product to be printed that needs to undergo laser transfer printing process; Selecting a product sample substrate and a product substrate to be printed according to the product type, and selecting corresponding performance test steps according to the index performance parameters to form a performance test process; In combination with the product type and the indicator performance parameter, the initial process parameters for processing the product sample substrate through the laser transfer printing process are retrieved from a preset laser printing process database; Based on the initial process parameters and in accordance with the laser transfer printing process, laser printing is performed on the product sample substrate to obtain a product sample to be printed; Performing a product performance test on the product sample to be printed according to the performance test process, and adjusting the initial process parameters to the optimal process parameters through the product performance test; Based on the optimal process parameters and in accordance with the laser transfer printing process, the laser printing process is performed on the substrate of the product to be printed.
[0006] Optionally, the performance testing steps include any one or more of a product clarity testing step, a product resolution testing step, a product gloss testing step, a product color stability testing step, a product adhesion testing step, a product light resistance testing step, a product water resistance testing step, a product chemical resistance testing step, a product mechanical property testing step, a product flatness testing step, a product environmental performance testing step and a product anti-counterfeiting performance testing step.
[0007] Optionally, performing a product performance test on the to-be-printed product sample according to the performance test process, and adjusting the initial process parameters to optimal process parameters through the product performance test comprises the following steps: Performing a performance test analysis step on the sample of the product to be printed, the performance test analysis step is as follows: Performing a product performance test on the product sample to be printed according to the performance test process, and obtaining measured performance parameters of the product performance test, wherein the measured performance parameters and the index performance parameters have the same number of parameter types; Calculating the parameter difference between the measured performance parameter and each parameter type of the index performance parameter respectively; If the parameter differences of all the parameter types do not exceed the preset difference threshold, the initial process parameters are used as the optimal process parameters, and the performance test analysis step is terminated.
[0008] Optionally, the method further comprises the following steps: If the parameter difference of at least one parameter type exceeds the difference threshold, a parameter optimization step is performed, and the parameter optimization step is as follows: Retrieving pre-stored parameter association information from the laser printing process database based on the index performance parameter, wherein the parameter association information includes association information between the index performance parameter and the process technology parameter; Marking the performance parameter corresponding to the parameter type as the performance parameter to be optimized, and marking the initial process parameter associated with the performance parameter to be optimized as the process parameter to be optimized according to the parameter association information; Selecting new process parameters as test process parameters within the parameter limit range of the process parameters to be optimized by using a binary search median approximation method; Based on all the test process parameters and all the unmarked initial process parameters, and in accordance with the laser transfer printing process, the substrate of the product to be printed is subjected to the laser printing process to obtain a test product sample; The performance test analysis step and the parameter optimization step are sequentially repeated on the test product sample to iteratively update the test process parameters until the differences between all the test process parameters and the corresponding index performance parameters do not exceed the difference threshold, and all the test process parameters and all the unmarked initial process parameters are taken as the optimal process parameters.
[0009] Optionally, performing a product performance test on the to-be-printed product sample according to the performance test process, and adjusting the initial process parameters to optimal process parameters through the product performance test comprises the following steps: Performing a product performance test on the product sample to be printed according to the performance test process, and obtaining measured performance parameters of the product performance test, wherein the measured performance parameters and the index performance parameters have the same number of parameter types; Acquiring printing processing data during the laser printing process; Building a parameter association analysis model based on a bidirectional long short-term memory network, using the parameter association analysis model to extract a first parameter association feature between the printing process data and the initial process parameters, and using the parameter association analysis model to extract a second parameter association feature between the printing process data and the index performance parameters; Building a performance parameter prediction model based on a graph neural network, combining the first parameter association feature and the second parameter association feature and using the performance parameter prediction model to predict the predicted performance parameters corresponding to the initial process parameters; By comprehensively analyzing the root mean square errors among the measured performance parameters, the predicted performance parameters and the index performance parameters, the initial process parameters are adjusted to the optimal process parameters.
[0010] Optionally, constructing a parameter association analysis model based on a bidirectional long short-term memory network, extracting a first parameter association feature between the printing process data and the initial process parameters using the parameter association analysis model, and extracting a second parameter association feature between the printing process data and the index performance parameter using the parameter association analysis model include the following steps: Constructing a parameter association analysis model based on a bidirectional long short-term memory network, the parameter association analysis model comprising a data preprocessing layer, a first association feature extraction layer, a second association feature extraction layer, and a first parameter feature association layer and a second parameter feature association layer based on the bidirectional long short-term memory network; Inputting the printing processing data into the parameter association analysis model, and performing data preprocessing on the printing processing data through the data preprocessing layer, wherein the data preprocessing includes data cleaning processing and data dimension reduction processing based on mutual information feature screening; Inputting the initial process parameters and the printing processing data after data preprocessing into the first parameter feature association layer and the first association feature extraction layer in sequence, and outputting the first parameter association feature between the printing processing data and the initial process parameters through the first association feature extraction layer; The index performance parameter and the printing processing data after data preprocessing are sequentially input into the second parameter feature association layer and the second association feature extraction layer, and the second parameter association feature between the printing processing data and the index performance parameter is output through the second association feature extraction layer.
[0011] Optionally, the step of constructing a performance parameter prediction model based on a graph neural network, combining the first parameter association feature and the second parameter association feature and using the performance parameter prediction model to predict the predicted performance parameter corresponding to the initial process parameter includes the following steps: Build a performance parameter prediction model based on graph neural network; Converting the initial process parameters, the printing processing data and the index performance parameters into graph structure data, and generating corresponding process parameter graph nodes, processing data graph nodes and index parameter graph nodes respectively; generating a first node edge between the process parameter graph node and the processing data graph node based on the first parameter association feature; generating a second node edge between the indicator parameter graph node and the processing data graph node based on the second parameter association feature to form a parameter association directed graph network; Acquiring historical laser printing data from the laser printing process database, wherein the historical laser printing data includes historical process parameters, historical printing processing data, and historical indicator performance parameters with parameter association tags; Using the historical laser printing data to complete the training process of the performance parameter prediction model; Inputting the parameter association directed graph network into the trained performance parameter prediction model, and aggregating second-order neighbor node features with the index parameter graph node as the target node through the performance parameter prediction model; A prediction result is output based on the second-order neighbor node characteristics, and the prediction result is converted into a prediction performance parameter corresponding to the initial process parameter.
[0012] Optionally, adjusting the initial process parameters to optimal process parameters by comprehensively analyzing the root mean square errors among the measured performance parameters, the predicted performance parameters and the index performance parameters comprises the following steps: Calculating a first root mean square error between the measured performance parameter and the predicted performance parameter; Determining whether the first root mean square error is less than a preset root mean square error threshold; If the first root mean square error is less than the root mean square error threshold, performing weighted average calculation on the measured performance parameter and the predicted performance parameter to obtain a test performance parameter; Calculating a second root mean square error between the test performance parameter and the index performance parameter; Determining whether the second root mean square error is less than the root mean square error threshold; If the second root mean square error is less than the root mean square error threshold, taking the initial process parameters as the optimal process parameters; If the second root mean square error is greater than or equal to the root mean square error threshold, the initial process parameters are iteratively updated based on the parameter association analysis model and the performance parameter prediction model and according to the second root mean square error until the second root mean square error is less than the root mean square error threshold, and the initial process parameters of the current iteration layer are used as the optimal process parameters.
[0013] Optionally, the method further comprises the following steps: If the first root mean square error is greater than or equal to the root mean square error threshold, model parameters of the parameter association analysis model and the performance parameter prediction model are iteratively updated based on the first root mean square error until the first root mean square error is less than the root mean square error threshold.
[0014] In a second aspect, the present invention further provides an intelligently controlled laser transverse drawing transfer printing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the intelligently controlled laser transverse drawing transfer printing method as described in the first aspect is implemented.
[0015] The beneficial effects of the present invention are: The technical solution provided by the present invention determines the product type of the product to be printed and the preset index performance parameters, selects the corresponding product sample substrate and the product substrate to be printed, and combines the performance testing steps to form a performance testing process, retrieves the initial process parameters from the preset laser printing process database, performs laser printing processing on the product sample substrate based on the initial process parameters, obtains the product sample to be printed, and performs performance testing on the product sample to be printed through the performance testing process, adjusts the initial process parameters to the optimal process parameters, and finally performs laser printing processing on the product substrate to be printed based on the optimal process parameters, thereby realizing the optimization of complex process parameters through a systematic and data-driven approach, reducing the time and errors of manual adjustment, improving production efficiency and the consistency and stability of product quality, and significantly reducing production costs and waste of resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 It is a schematic diagram of the process of the intelligently controlled laser transverse wire drawing transfer printing method in one embodiment of the present application.
[0017] Figure 2 FIG. 1 is a schematic diagram of a laser transfer printing process in one embodiment of the present application.
[0018] Figure 3 A schematic diagram of a model framework of a parameter association analysis model in one implementation manner of the present application. DETAILED DESCRIPTION
[0019] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.
[0020] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.
[0021] Figure 1 FIG. 1 is a flow chart of a method for intelligently controlling laser lateral wire drawing transfer printing in one embodiment. It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps in the above process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps. The intelligently controlled laser transverse drawing transfer printing method disclosed in the present invention is applied to the fully automatic laser transverse drawing transfer printing process, such as Figure 1 As shown, the intelligently controlled laser transverse wire drawing transfer printing method specifically includes the following steps: S101. Determine the product type of the product to be printed that needs to undergo a laser transfer printing process and preset index performance parameters.
[0022] Among them, in order to determine the product type of the product to be printed that needs to undergo the laser transfer printing process, the product needs to be classified first. This can be achieved by analyzing the product's characteristics such as purpose, material, shape and size. For example, if the product is a plastic film used for packaging, then its product type can be classified as a plastic packaging material. After determining the product type, the product's index performance parameters need to be preset. These parameters include but are not limited to the product's clarity, resolution, gloss, color stability, adhesion, light resistance, water resistance, chemical resistance, mechanical properties, flatness, environmental protection performance and anti-counterfeiting performance. Each index performance parameter needs to have a specific numerical value or descriptive standard, for example, clarity can be measured in DPI (dots per inch), and gloss can be expressed in gloss units measured by a gloss meter. In order to determine these parameters, industry standards, customer needs and specific application scenarios of the product can be referred to.
[0023] like Figure 2 As shown in the figure, the laser transfer printing process is as follows: first, design a plane image, and then, based on the plane image design, perform photolithography design of suspended graphics elements to generate graphics with three-dimensional effects. The design of the suspended graphics elements is transferred to the photoresist through photolithography technology. Photolithography is the process of transferring a pattern to a substrate using a photochemical reaction. The photoresist is developed to reveal the pattern after photolithography. Development is to dissolve the unexposed photoresist with a chemical solution, leaving the exposed pattern. The developed pattern is transferred to a nickel roller through an electroforming process, and then the electroforming nickel roller is used to press the graphics elements on the mold. Electroforming is the process of depositing metal on the mold surface using an electrochemical reaction.
[0024] PET base film is a polymer material commonly used in printing, with good mechanical properties and chemical stability. The coating material can be various functional coatings, such as photosensitive coatings, conductive coatings, etc. The PET base film is coated so that it can better accept the image element molding. Preliminary heating treatment is carried out under constant temperature conditions to make the coating material evenly distributed and solidified. The heated PET base film is heat-dissipated to stabilize the coating performance. The prepared image element die is molded on the PET base film. The molded PET base film is further heated under constant temperature conditions to ensure the stability and adhesion of the image element. The heated PET base film is heat-dissipated to stabilize the performance of the image element. The treated PET base film is then guided to the next process through a roller cooling device.
[0025] In the next process, the PET base film is aluminized to increase its reflective properties and optical effects. A glue layer is applied to the aluminized PET base film to facilitate the subsequent paper covering process. The paper is covered on the glue-coated PET base film to form the final printed product. The paper-covered product is heated under constant temperature conditions to ensure the curing and adhesion of the glue layer. The heated product is heat-dissipated to stabilize its performance. After completing all the above processes, a graphic element product with naked-eye 3D effect is obtained. The completed 3D graphic element product is rolled up for easy storage and transportation.
[0026] S102. Select product sample substrates and to-be-printed product substrates according to product types, and select corresponding performance test steps according to index performance parameters to form a performance test process.
[0027] Among them, according to the determined product type, a suitable product sample substrate and a product substrate to be printed are selected. The product sample substrate is a material used to test and adjust process parameters, while the product substrate to be printed is a material used in actual production. For example, if the product type is a plastic film, a film of the same material but different thickness can be selected as a sample substrate to test the effect of different thicknesses on the printing effect. Next, the corresponding performance test steps are selected according to the preset index performance parameters. The performance test steps include any one or more of the product clarity test step, the product resolution test step, the product gloss test step, the product color stability test step, the product adhesion test step, the product light resistance test step, the product water resistance test step, the product chemical resistance test step, the product mechanical performance test step, the product flatness test step, the product environmental protection performance test step and the product anti-counterfeiting performance test step.
[0028] S103. Based on the product type and the index performance parameters, the initial process parameters of the product sample substrate processed by the laser transfer printing process are retrieved from the preset laser printing process database.
[0029] First, a database containing various laser printing process parameters needs to be established. The database should include process parameters corresponding to different product types and performance parameters, such as laser power, scanning speed, focal length, exposure time, etc. According to the product type and index performance parameters determined in the previous step, the corresponding initial process parameters are retrieved from the database. For example, if the product type is a plastic film and the clarity requirement is high, process parameters with high laser power and slow scanning speed can be selected. After retrieving these parameters, appropriate adjustments need to be made according to the specific product sample substrate to ensure that the process parameters can adapt to changes in actual production.
[0030] S104. Based on the initial process parameters and in accordance with the laser transfer printing process, laser printing is performed on the product sample substrate to obtain a product sample to be printed.
[0031] S105. Perform product performance testing on the printed product samples according to the performance testing process, and adjust the initial process parameters to the optimal process parameters through the product performance testing.
[0032] Among them, according to the performance test steps selected above, the printed product samples are comprehensively tested for performance. For example, the clarity of the printed pattern can be observed using a microscope, the gloss meter can be used to measure the gloss of the product, and the colorimeter can be used to detect color stability. Each test step requires specific operating methods and standards. For example, the clarity test can be evaluated by measuring the minimum line width of the pattern, and the gloss test can be performed by measuring the intensity of reflected light. After completing all tests, the initial process parameters are adjusted according to the test results.
[0033] S106. Perform laser printing on the substrate of the product to be printed based on the optimal process parameters and in accordance with the laser transfer printing process.
[0034] In one embodiment, performing a product performance test on a printed product sample according to a performance test process, and adjusting the initial process parameters to the optimal process parameters through the product performance test includes the following steps: The performance test analysis steps are performed on the printed product samples. The performance test analysis steps are as follows: Perform product performance testing on the printed product samples according to the performance testing process, and obtain the measured performance parameters of the product performance testing, where the measured performance parameters and the index performance parameters have the same number of parameter types; Calculate the parameter differences of each parameter type between the measured performance parameters and the index performance parameters respectively; If the parameter differences of all parameter types do not exceed the preset difference threshold, the initial process parameters are used as the optimal process parameters, and the performance test analysis step is terminated.
[0035] In this embodiment, the product performance test is performed on the printed product samples according to the performance test process, and the measured performance parameters of the product performance test are obtained. The performance test process is to perform a series of tests and measurements on the printed product samples according to the pre-established test standards and steps to evaluate whether their performance meets expectations. The measured performance parameters refer to the numerical data obtained during the actual test process, which reflect the performance of the printed product samples in various performance indicators. The number of types of measured performance parameters should be the same as the index performance parameters to ensure that a comprehensive comparative analysis can be performed. Then, the parameter difference between the measured performance parameters and each parameter type of the index performance parameters is calculated respectively. The parameter difference is an important indicator to measure the difference between the measured performance and the expected performance. By calculating the difference of each parameter type, the deviation between the actual performance of each performance indicator and the expected target can be clearly defined.
[0036] Then determine whether the parameter differences of all parameter types do not exceed the preset difference threshold. The preset difference threshold is an allowable range set according to product performance requirements and process control standards, reflecting the maximum acceptable deviation of each performance parameter. The specific implementation method can be achieved by comparing the parameter difference of each parameter type with the preset difference threshold. If the parameter differences of all parameter types do not exceed the preset difference threshold, it means that the performance of the product sample to be printed is within an acceptable range; otherwise, it means that the sample has deviations in certain performance indicators and needs to further adjust the process parameters. The implementation effect of this step is to evaluate the overall performance of the product sample to be printed, determine whether it meets the expected requirements, and provide a basis for the optimization of process parameters.
[0037] If the parameter differences of all parameter types do not exceed the preset difference threshold, the initial process parameters are used as the optimal process parameters and the performance test analysis step is terminated. The specific implementation method can record the current initial process parameters and mark them as optimal parameters. The initial process parameters refer to the process control parameters set before the performance test, which determine the manufacturing process and performance of the product. Using them as the optimal process parameters means that the current process control solution has achieved the expected performance requirements and does not require further adjustment.
[0038] In this embodiment, if the parameter difference of at least one parameter type exceeds the difference threshold, a parameter optimization step is performed, and the parameter optimization step is as follows: Retrieving pre-stored parameter association information from a laser printing process database based on the index performance parameters, the parameter association information including association information between the index performance parameters and the process parameters; Marking the performance parameters of the corresponding parameter types as performance parameters to be optimized, and marking the initial process parameters associated with the performance parameters to be optimized as process parameters to be optimized according to the parameter association information; Using the binary search median approximation method, new process parameters are selected as test process parameters within the parameter limit range of the process parameters to be optimized; Based on all the test process parameters and all the unmarked initial process parameters, laser printing is performed on the substrate of the product to be printed according to the laser transfer printing process to obtain a test product sample; Repeat the performance test analysis steps and parameter optimization steps on the test product samples in sequence to iteratively update the test process parameters until the differences between all test process parameters and the corresponding indicator performance parameters do not exceed the difference threshold, and take all test process parameters and all unmarked initial process parameters as the optimal process parameters.
[0039] In this embodiment, first, by querying the laser printing process database, all historical data related to the current indicator performance parameters are retrieved. These data include the performance of each performance parameter under different process conditions, as well as the corresponding process parameters. The implementation method can obtain these data through database query statements such as SQL. The performance parameters of the corresponding parameter types are marked as performance parameters to be optimized, and the initial process parameters associated with the performance parameters to be optimized are marked as process parameters to be optimized according to the parameter association information. The specific implementation principle is to determine which performance parameters need to be optimized by analyzing the parameter association information, and find the process parameters related to these performance parameters.
[0040] Then, the binary search median approximation method is used to select new process parameters as test process parameters within the parameter limit range of the process parameters to be optimized. The specific implementation principle is to gradually approach the optimal value within the given parameter range through the binary search algorithm. The binary search median approximation method is an efficient numerical optimization method that gradually approaches the target value by continuously reducing the search range by half. The implementation method can be achieved through the following steps: First, determine the parameter range [L, U] of the process parameters to be optimized, where L is the lower limit and U is the upper limit; then, calculate the median M=(L+U) / 2; then, test according to the median M, if the test results meet the requirements, use M as the new process parameter; otherwise, adjust the parameter range according to the test results and continue the binary search. The effect of this step is to quickly find a process parameter close to the optimal one and improve the optimization efficiency.
[0041] Apply all determined process parameters to the actual laser printing process to generate test product samples. Then, the performance test analysis steps and parameter optimization steps need to be repeated on the test product samples in sequence to iteratively update the test process parameters until the difference between all test process parameters and the corresponding index performance parameters does not exceed the difference threshold, and all test process parameters and all unmarked initial process parameters are taken as the optimal process parameters. The specific implementation principle is to adjust the process parameters continuously by repeatedly performing performance tests and parameter optimization until all performance parameters meet the expected requirements. The implementation method can be achieved through the following steps: first, perform performance tests on the test product samples to obtain the measured performance parameters; then, calculate the difference between the measured performance parameters and the index performance parameters; then, adjust the process parameters to be optimized according to the difference, and select new test process parameters; repeat the above steps until all differences do not exceed the preset threshold. The effect of this step is to finally determine a set of optimal process parameters to ensure that the product performance meets the expected requirements.
[0042] In one embodiment, performing a product performance test on a printed product sample according to a performance test process, and adjusting the initial process parameters to the optimal process parameters through the product performance test includes the following steps: Perform product performance testing on the printed product samples according to the performance testing process, and obtain the measured performance parameters of the product performance testing, where the measured performance parameters and the index performance parameters have the same number of parameter types; Acquire printing processing data during laser printing processing; A parameter association analysis model is constructed based on a bidirectional long short-term memory network, and the first parameter association characteristics between the printing processing data and the initial process parameters are extracted using the parameter association analysis model, and the second parameter association characteristics between the printing processing data and the index performance parameters are extracted using the parameter association analysis model; A performance parameter prediction model is constructed based on a graph neural network, and the predicted performance parameters corresponding to the initial process parameters are predicted using the performance parameter prediction model by combining the first parameter association feature and the second parameter association feature; By comprehensively analyzing the root mean square errors among the measured performance parameters, predicted performance parameters and index performance parameters, the initial process parameters are adjusted to the optimal process parameters.
[0043] In this embodiment, first, the product performance test is performed on the sample of the product to be printed to obtain the measured performance parameters of the product performance test. In the specific implementation process, the sample of the product to be printed is subjected to a series of performance tests under specific conditions, including but not limited to wear resistance, weather resistance, adhesion, optical properties and mechanical strength. These tests are carried out by standardized test methods and equipment, and each test is carried out in strict accordance with relevant standards and specifications to ensure the accuracy and reliability of the data. The test results are recorded as measured performance parameters, which are the same as the number of types of index performance parameters, which is convenient for subsequent comparative analysis. While obtaining the measured performance parameters, the printing processing data in the laser printing process is collected. The printing processing data includes but is not limited to printing speed, temperature, pressure, coating thickness, aluminum coating thickness, molding depth, etc. By installing sensors and data recording systems on the printing equipment, these parameters are monitored and recorded in real time. The data recording system can accurately capture the parameter changes of each processing step and store them in a database. These data not only reflect the process conditions of the printing process, but also provide a basis for analyzing printing quality and performance.
[0044] A parameter association analysis model is constructed based on a bidirectional long short-term memory network (Bi-LSTM) to analyze the relationship between printing processing data and initial process parameters. Bi-LSTM is a deep learning model that can capture long-term and short-term dependencies in time series data. By taking printing processing data as input and initial process parameters as output, the Bi-LSTM model is trained to enable it to learn and extract the first parameter association features between printing processing data and initial process parameters. Similarly, the same Bi-LSTM model is used to analyze the relationship between printing processing data and indicator performance parameters and extract the second parameter association features. During the training process, the back propagation algorithm is used to optimize the model parameters, minimize the prediction error, and improve the prediction accuracy of the model.
[0045] Based on the graph neural network (GNN), a performance parameter prediction model is constructed, and the predicted performance parameters corresponding to the initial process parameters are predicted by combining the first parameter association feature and the second parameter association feature. GNN is a deep learning model that can process graph structure data and is suitable for capturing complex parameter relationships. The first parameter association feature and the second parameter association feature are used as the features of the graph nodes to construct the graph structure and input the GNN model for training. During the training process, graph convolution operations and pooling operations are used to extract useful information from the graph structure and generate predicted performance parameters. Through repeated training and verification, the GNN model parameters are optimized so that it can accurately predict the performance parameters corresponding to the initial process parameters.
[0046] Finally, by comprehensively analyzing the root mean square error between the measured performance parameters, predicted performance parameters, and index performance parameters, the initial process parameters are adjusted to the optimal process parameters. The root mean square error (RMSE) is a standard indicator for measuring the difference between the predicted value and the actual value. The RMSE between the measured performance parameters and the predicted performance parameters is calculated to evaluate the accuracy of the prediction model. At the same time, the RMSE between the measured performance parameters and the index performance parameters is calculated to evaluate the performance of the current process parameters. By iteratively adjusting the initial process parameters, the RMSE between the measured performance parameters and the index performance parameters is minimized, and the optimal process parameters are finally determined to ensure that the product performance meets the expected standards.
[0047] In one embodiment, constructing a parameter association analysis model based on a bidirectional long short-term memory network, extracting a first parameter association feature between printing processing data and initial process parameters using the parameter association analysis model, and extracting a second parameter association feature between printing processing data and index performance parameters using the parameter association analysis model include the following steps: Constructing a parameter association analysis model based on a bidirectional long short-term memory network, the parameter association analysis model comprising a data preprocessing layer, a first association feature extraction layer, a second association feature extraction layer, and a first parameter feature association layer and a second parameter feature association layer based on a bidirectional long short-term memory network; Input the printing processing data into the parameter association analysis model, and perform data preprocessing on the printing processing data through the data preprocessing layer, the data preprocessing includes data cleaning processing and data dimension reduction processing based on mutual information feature screening; Inputting the initial process parameters and the printing processing data after data preprocessing into the first parameter feature association layer and the first association feature extraction layer in sequence, and outputting the first parameter association feature between the printing processing data and the initial process parameters through the first association feature extraction layer; The index performance parameters and the printing processing data after data preprocessing are sequentially input into the second parameter feature association layer and the second association feature extraction layer, and the second parameter association feature between the printing processing data and the index performance parameters is output through the second association feature extraction layer.
[0048] In this embodiment, refer to Figure 3, in actual working conditions, some processing data are missing due to equipment failure and other reasons, and the measurement units of complex product printing processing data are many and the differences are large. In order to eliminate the impact of dimension and data missing. The printing processing data needs to be preprocessed. The data preprocessing layer optimizes the data through data cleaning and data dimensionality reduction based on mutual information feature screening. The goal of data cleaning is to remove or repair noise and outliers in the data to ensure the accuracy and consistency of the data. Data cleaning can include steps such as removing duplicate data, filling missing values, and correcting data entry errors. To achieve these steps, a variety of techniques can be applied, such as filling missing values with the mean or median, or using machine learning algorithms to predict missing values. Data dimensionality reduction reduces the dimension of the data through feature screening based on mutual information. Mutual information is a statistic that measures the mutual dependence between two variables, which can help identify the features that have the greatest impact on the target variable. By calculating the mutual information value between each feature and the target variable, the most relevant features can be screened out, thereby reducing the data dimension and reducing the computational complexity while retaining important information. Ultimately, the data after data cleaning and dimensionality reduction will be more concise and representative, providing a high-quality data foundation for subsequent feature extraction and association analysis.
[0049] The initial process parameters and pre-processed printing processing data will be input into the first parameter feature association layer and the first association feature extraction layer in sequence. The first parameter feature association layer processes the input data based on the bidirectional long short-term memory network (Bi-LSTM). Bi-LSTM is a deep learning model that can capture the long-term and short-term dependencies in sequence data and has the ability to process sequence information simultaneously in both forward and backward directions. Specifically, Bi-LSTM processes the forward and reverse sequences of data respectively through two LSTM units in the forward and backward directions, so as to more comprehensively understand the context information of the data. In the first parameter feature association layer, the initial process parameters and pre-processed printing processing data will be combined into a multi-dimensional time series and input into the Bi-LSTM network. The Bi-LSTM network controls the flow of information through its gating mechanism (including input gate, forget gate and output gate), captures the long-term and short-term dependencies in the data, and generates high-dimensional hidden state representations. Subsequently, these hidden state representations will be input into the first association feature extraction layer. The function of the first association feature extraction layer is to extract features that are highly correlated with the initial process parameters from the hidden states generated by Bi-LSTM. This can be achieved by applying neural network layers such as fully connected layers or convolutional layers to further process and extract key features. Ultimately, the output of the first correlation feature extraction layer will be the first parameter correlation features between the printing process data and the initial process parameters, which can be used to further analyze and optimize the printing process.
[0050] Similar to the first parameter feature association layer, the second parameter feature association layer also processes input data based on the Bi-LSTM network. The difference is that the second parameter feature association layer aims to capture the association features between printing process data and index performance parameters. Index performance parameters are usually key indicators used to measure the quality and production efficiency of printed products, such as color consistency, printing speed, scrap rate, etc. By combining these index performance parameters with the pre-processed printing process data into a multidimensional time series and inputting them into the Bi-LSTM network, the network can capture the complex dependencies in the data and generate corresponding hidden state representations. Subsequently, these hidden state representations will be input into the second association feature extraction layer. The second association feature extraction layer extracts features that are highly correlated with the index performance parameters from the hidden state by applying appropriate neural network layers. These features can include key control variables in the printing process, major factors affecting quality, etc. Ultimately, the output of the second association feature extraction layer will be the second parameter association features between printing process data and index performance parameters. These features can help identify key factors affecting printing quality and efficiency, and provide valuable reference information to optimize production processes and improve product quality.
[0051] In summary, through the collaborative work of the data preprocessing layer, the first parameter feature association layer and the first association feature extraction layer, as well as the second parameter feature association layer and the second association feature extraction layer, the complex association relationship between the printing processing data and the initial process parameters and the index performance parameters can be effectively analyzed. The data preprocessing layer ensures the quality and simplicity of the input data through cleaning and dimensionality reduction. The first parameter feature association layer and the first association feature extraction layer use the Bi-LSTM network to capture the long-term and short-term dependencies between the printing processing data and the initial process parameters, and extract key association features. The second parameter feature association layer and the second association feature extraction layer focus on analyzing the association features between the printing processing data and the index performance parameters, helping to identify the key factors affecting printing quality and efficiency.
[0052] In one embodiment, a performance parameter prediction model is constructed based on a graph neural network, and the predicted performance parameters corresponding to the initial process parameters are predicted by using the performance parameter prediction model in combination with the first parameter association feature and the second parameter association feature, including the following steps: Build a performance parameter prediction model based on graph neural network; Convert the initial process parameters, printing processing data and index performance parameters into graph structure data, and generate corresponding process parameter graph nodes, processing data graph nodes and index parameter graph nodes respectively; Generate a first node edge between a process parameter graph node and a processing data graph node based on a first parameter association feature; generating a second node edge between the indicator parameter graph node and the processing data graph node based on the second parameter association feature to form a parameter association directed graph network; Acquire historical laser printing data from a laser printing process database, the historical laser printing data including historical process parameters with parameter association tags, historical printing processing data, and historical indicator performance parameters; Use historical laser printing data to complete the training process of the performance parameter prediction model; The parameter association directed graph network is input into the trained performance parameter prediction model, and the second-order neighbor node features are aggregated by the performance parameter prediction model with the index parameter graph node as the target node; The prediction results are output based on the second-order neighbor node characteristics, and the prediction results are converted into prediction performance parameters corresponding to the initial process parameters.
[0053] In this embodiment, the initial process parameters, printing processing data and index performance parameters are first converted into graph structure data, and corresponding process parameter graph nodes, processing data graph nodes and index parameter graph nodes are generated respectively. Graph structure data is a structured data form that can represent entities and their relationships, where nodes represent entities and edges represent the relationships between entities. In this step, the initial process parameters, printing processing data and index performance parameters are regarded as different types of entities and converted into graph nodes respectively. The specific implementation method can be implemented by defining node feature vectors for each parameter type. For example, the initial process parameters may include features such as temperature, pressure, speed, etc., and each feature value may be used as a node attribute. The printing processing data may include time series data, such as printing speed, ink usage, etc., which can be converted into node feature vectors by feature extraction methods. The index performance parameters are usually the final quality indicators, such as color consistency, printing accuracy, etc., which can be directly used as node attributes. In this way, all parameters are converted into nodes in the graph structure, laying the foundation for subsequent graph neural network processing.
[0054] Next, the first node edge between the process parameter graph node and the processing data graph node is generated based on the first parameter association feature. The first parameter association feature is extracted by the aforementioned bidirectional long short-term memory network (Bi-LSTM), which reflects the association relationship between the initial process process parameters and the printing processing data. In the graph structure, these association relationships can be represented by edges. A specific implementation method can determine the weight of the edge by calculating the similarity or correlation between the process parameter node and the processing data node. For example, methods such as cosine similarity and Pearson correlation coefficient can be used to calculate the similarity values between node feature vectors, and these values are used as edge weights. The direction of the edge can be determined based on the causality of the association relationship, that is, from the process parameter node to the processing data node. In this way, the first node edge between the process parameter graph node and the processing data graph node is generated, forming a partial parameter association directed graph network.
[0055] Similarly, a second node edge between the indicator parameter graph node and the processing data graph node is generated based on the second parameter association feature. The second parameter association feature reflects the association relationship between the printing processing data and the indicator performance parameters, and these features are also extracted through the Bi-LSTM network. In the graph structure, these associations can be represented by edges. The specific implementation method can be similar to the generation of the first node edge, and the weight of the edge is determined by calculating the similarity or correlation between the processing data node and the indicator parameter node. The direction of the edge can be determined according to the causality of the association relationship, that is, from the processing data node to the indicator parameter node. In this way, the second node edge between the indicator parameter graph node and the processing data graph node is generated, further improving the parameter association directed graph network.
[0056] Then, historical laser printing data is obtained from the laser printing process database. The historical laser printing data includes historical process parameters with parameter association labels, historical printing processing data, and historical indicator performance parameters. Historical data is an important basis for model training. Through these data, complex correlations between parameters can be learned. The specific implementation method can obtain the required historical data by querying the database and ensure the integrity and accuracy of the data. Historical data usually includes multiple samples, each of which contains complete process parameters, processing data, and performance indicators. In order to improve the quality of the data, the historical data can be preprocessed, such as removing noise, processing missing values, etc.
[0057] Then, the training process of the performance parameter prediction model is completed using the historical laser printing data. The performance parameter prediction model is built based on a graph neural network (GNN), which is a deep learning model that can process graph structured data. The specific implementation method can be achieved by defining the architecture of the graph neural network, including an input layer, a hidden layer, and an output layer, and selecting appropriate activation functions and loss functions. The input layer receives graph structured data, including node features and edge weights. The hidden layer aggregates the neighbor information of the node through graph convolution operations to capture the complex associations between nodes. The output layer generates a prediction result, which is usually a feature vector of the node. The goal of model training is to minimize the error between the prediction result and the actual performance parameter. The specific implementation method can adjust the model parameters through a back-propagation algorithm. In this way, the performance parameter prediction model is trained using historical laser printing data, so that it can accurately predict the performance parameters of the printing process.
[0058] The parameter association directed graph network is input into the trained performance parameter prediction model, and the performance parameter prediction model is used to aggregate the second-order neighbor node features with the index parameter graph node as the target node. An important feature of the graph neural network is that it can aggregate the neighbor information of the node, thereby capturing the complex association relationship between the nodes. The specific implementation method can be implemented through multi-layer graph convolution operations, and each layer of graph convolution aggregates the features of the neighbor nodes once. With the index parameter graph node as the target node, through multi-layer graph convolution operations, the features of the second-order neighbor nodes, that is, the features of the neighbor nodes of the target node and its neighbor nodes, can be aggregated. In this way, the complex association relationship in the parameter association directed graph network can be captured, providing rich information for the accurate prediction of performance parameters. The output layer of the graph neural network generates the feature vector of the target node, which can be further processed by the neural network layer such as the fully connected layer to generate the final prediction result. The specific implementation method can measure the difference between the prediction result and the actual performance parameter by defining an appropriate loss function, such as the mean square error (MSE), and optimize the model parameters through the back propagation algorithm. The final prediction result can be converted into the predicted performance parameter corresponding to the initial process parameters through the decoder.
[0059] In one embodiment, adjusting the initial process parameters to the optimal process parameters by comprehensively analyzing the root mean square errors among the measured performance parameters, the predicted performance parameters, and the index performance parameters comprises the following steps: calculating a first root mean square error between the measured performance parameter and the predicted performance parameter; Determining whether the first root mean square error is less than a preset root mean square error threshold; If the first root mean square error is greater than or equal to a root mean square error threshold, iteratively updating model parameters of the parameter association analysis model and the performance parameter prediction model based on the first root mean square error until the first root mean square error is less than the root mean square error threshold; If the first root mean square error is less than the root mean square error threshold, a weighted average calculation is performed on the measured performance parameter and the predicted performance parameter to obtain the test performance parameter; Calculate the second root mean square error between the test performance parameter and the index performance parameter; Determine whether the second root mean square error is less than a root mean square error threshold; If the second root mean square error is less than the root mean square error threshold, the initial process parameters are used as the optimal process parameters; If the second root mean square error is greater than or equal to the root mean square error threshold, the initial process parameters are iteratively updated based on the parameter association analysis model and the performance parameter prediction model and according to the second root mean square error until the second root mean square error is less than the root mean square error threshold, and the initial process parameters of the current iteration layer are used as the optimal process parameters.
[0060] In this embodiment, the first root mean square error between the measured performance parameter and the predicted performance parameter is first calculated. The root mean square error (RMSE) is a commonly used indicator to measure the accuracy of the prediction model, reflecting the difference between the predicted value and the true value. The first root mean square error can be obtained by calculating the mean of the sum of squared differences of all samples and then taking the square root. Next, it is determined whether the first root mean square error is less than a preset root mean square error threshold. The preset root mean square error threshold is a predefined error tolerance used to evaluate the prediction performance of the model.
[0061] If the first root mean square error is less than the root mean square error threshold, it means that the prediction accuracy of the model is within an acceptable range. At this time, it is necessary to perform a weighted average calculation on the measured performance parameters and the predicted performance parameters to obtain the test performance parameters. Weighted average is a method of combining multiple data sources to balance their influence by assigning different weights to each data source. At this time, it is necessary to calculate the second root mean square error between the test performance parameters and the index performance parameters. The calculation method of the second root mean square error is similar to that of the first root mean square error, except that the test performance parameters and the index performance parameters are used for calculation. Next, determine whether the second root mean square error is less than the root mean square error threshold. If the second root mean square error is less than the root mean square error threshold, it means that the difference between the test performance parameters and the index performance parameters is within an acceptable range, and the initial process parameters can be used as the optimal process parameters; otherwise, it means that the initial process parameters need to be further optimized. The process of further optimizing the initial process parameters is as follows: Based on the parameter association analysis model and the performance parameter prediction model and according to the second root mean square error, the initial process parameters are iteratively updated until the second root mean square error is less than the root mean square error threshold, and the initial process parameters of the current iteration layer are used as the optimal process parameters. The specific implementation method can be achieved through the following steps: First, based on the parameter association analysis model and the performance parameter prediction model, new initial process parameters are generated; then, the aforementioned error calculation and judgment steps are repeated to gradually optimize the process parameters. Through this iterative update method, the optimal process parameters that meet the performance requirements can be gradually approached. The implementation effect of this step is to finally find the optimal process parameters that meet the performance requirements through multiple iterative optimizations.
[0062] If the first root mean square error is greater than or equal to the root mean square error threshold, the model parameters of the parameter association analysis model and the performance parameter prediction model are iteratively updated based on the first root mean square error until the first root mean square error is less than the root mean square error threshold. The specific implementation method can be achieved through the following steps: First, based on the current first root mean square error, adjust the parameters of the parameter association analysis model and the performance parameter prediction model; then, retrain the model and calculate the new first root mean square error. Through this iterative update method, the prediction accuracy of the model can be gradually optimized. The implementation effect of this step is that through multiple iterative optimizations, a parameter association analysis model and a performance parameter prediction model that meet the performance requirements are finally obtained, providing a reliable prediction tool for the optimization of process parameters.
[0063] The present invention also discloses an intelligently controlled laser transverse wire drawing transfer printing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the intelligently controlled laser transverse wire drawing transfer printing method described in any one of the above-mentioned embodiments is implemented.
[0064] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.
[0065] Among them, the memory can be an internal storage unit of a computer device, such as a hard disk or memory of a computer device, or an external storage device of a computer device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the computer device, and the memory can also be a combination of an internal storage unit and an external storage device of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.
[0066] A person skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of protection of the present application is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present application as above, which are not provided in detail for the sake of simplicity.
[0067] One or more embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present application should be included in the protection scope of the present application.
Claims
1. An intelligently controlled laser transverse wire drawing transfer printing method, characterized in that: The steps include: Determine the product type and preset performance parameters of the product to be printed that needs to undergo laser transfer printing process; Selecting a product sample substrate and a product substrate to be printed according to the product type, and selecting corresponding performance test steps according to the index performance parameters to form a performance test process; In combination with the product type and the index performance parameter, the initial process parameters for processing the product sample substrate through the laser transfer printing process are retrieved from a preset laser printing process database; Based on the initial process parameters and in accordance with the laser transfer printing process, laser printing is performed on the product sample substrate to obtain a product sample to be printed; Performing a product performance test on the product sample to be printed according to the performance test process, and adjusting the initial process parameters to the optimal process parameters through the product performance test; Based on the optimal process parameters and in accordance with the laser transfer printing process, the laser printing process is performed on the substrate of the product to be printed.
2. The intelligent controlled laser lateral drawing transfer printing method according to claim 1, characterized in that: The performance testing steps include any one or more of a product clarity testing step, a product resolution testing step, a product gloss testing step, a product color stability testing step, a product adhesion testing step, a product light resistance testing step, a product water resistance testing step, a product chemical resistance testing step, a product mechanical performance testing step, a product flatness testing step, a product environmental performance testing step and a product anti-counterfeiting performance testing step.
3. The intelligent controlled laser lateral drawing transfer printing method according to claim 2, characterized in that: The steps of performing a product performance test on the product sample to be printed according to the performance test process and adjusting the initial process parameters to the optimal process parameters through the product performance test include the following steps: Performing a performance test analysis step on the sample of the product to be printed, the performance test analysis step is as follows: Performing a product performance test on the product sample to be printed according to the performance test process, and obtaining measured performance parameters of the product performance test, wherein the measured performance parameters and the index performance parameters have the same number of parameter types; Calculating the parameter difference between the measured performance parameter and each parameter type of the index performance parameter respectively; If the parameter differences of all the parameter types do not exceed the preset difference threshold, the initial process parameters are used as the optimal process parameters, and the performance test analysis step is terminated.
4. The intelligent controlled laser lateral drawing transfer printing method according to claim 3, characterized in that: The method further comprises the steps of: If the parameter difference of at least one parameter type exceeds the difference threshold, a parameter optimization step is performed, and the parameter optimization step is as follows: Retrieving pre-stored parameter association information from the laser printing process database based on the index performance parameter, wherein the parameter association information includes association information between the index performance parameter and the process technology parameter; Marking the performance parameter corresponding to the parameter type as the performance parameter to be optimized, and marking the initial process parameter associated with the performance parameter to be optimized as the process parameter to be optimized according to the parameter association information; Selecting new process parameters as test process parameters within the parameter limit range of the process parameters to be optimized by using a binary search median approximation method; Based on all the test process parameters and all the unmarked initial process parameters, and in accordance with the laser transfer printing process, the substrate of the product to be printed is subjected to the laser printing process to obtain a test product sample; The performance test analysis step and the parameter optimization step are sequentially repeated on the test product sample to iteratively update the test process parameters until the differences between all the test process parameters and the corresponding index performance parameters do not exceed the difference threshold, and all the test process parameters and all the unmarked initial process parameters are taken as the optimal process parameters.
5. The intelligent controlled laser lateral drawing transfer printing method according to claim 1, characterized in that: The steps of performing a product performance test on the product sample to be printed according to the performance test process and adjusting the initial process parameters to the optimal process parameters through the product performance test include the following steps: Performing a product performance test on the product sample to be printed according to the performance test process, and obtaining measured performance parameters of the product performance test, wherein the measured performance parameters and the index performance parameters have the same number of parameter types; Acquiring printing processing data during the laser printing process; Building a parameter association analysis model based on a bidirectional long short-term memory network, using the parameter association analysis model to extract a first parameter association feature between the printing process data and the initial process parameters, and using the parameter association analysis model to extract a second parameter association feature between the printing process data and the index performance parameters; Building a performance parameter prediction model based on a graph neural network, combining the first parameter association feature and the second parameter association feature and using the performance parameter prediction model to predict the predicted performance parameters corresponding to the initial process parameters; By comprehensively analyzing the root mean square errors among the measured performance parameters, the predicted performance parameters and the index performance parameters, the initial process parameters are adjusted to the optimal process parameters.
6. The intelligent controlled laser lateral drawing transfer printing method according to claim 5, characterized in that: The method of constructing a parameter association analysis model based on a bidirectional long short-term memory network, extracting a first parameter association feature between the printing process data and the initial process parameters by using the parameter association analysis model, and extracting a second parameter association feature between the printing process data and the index performance parameter by using the parameter association analysis model comprises the following steps: Constructing a parameter association analysis model based on a bidirectional long short-term memory network, the parameter association analysis model comprising a data preprocessing layer, a first association feature extraction layer, a second association feature extraction layer, and a first parameter feature association layer and a second parameter feature association layer based on the bidirectional long short-term memory network; Inputting the printing processing data into the parameter association analysis model, and performing data preprocessing on the printing processing data through the data preprocessing layer, wherein the data preprocessing includes data cleaning processing and data dimension reduction processing based on mutual information feature screening; Inputting the initial process parameters and the printing processing data after data preprocessing into the first parameter feature association layer and the first association feature extraction layer in sequence, and outputting the first parameter association feature between the printing processing data and the initial process parameters through the first association feature extraction layer; The index performance parameter and the printing processing data after data preprocessing are sequentially input into the second parameter feature association layer and the second association feature extraction layer, and the second parameter association feature between the printing processing data and the index performance parameter is output through the second association feature extraction layer.
7. The intelligent controlled laser lateral drawing transfer printing method according to claim 5, characterized in that: The step of constructing a performance parameter prediction model based on a graph neural network, combining the first parameter association feature and the second parameter association feature and using the performance parameter prediction model to predict the predicted performance parameter corresponding to the initial process parameter comprises the following steps: Build a performance parameter prediction model based on graph neural network; Converting the initial process parameters, the printing processing data and the index performance parameters into graph structure data, and generating corresponding process parameter graph nodes, processing data graph nodes and index parameter graph nodes respectively; generating a first node edge between the process parameter graph node and the processing data graph node based on the first parameter association feature; generating a second node edge between the indicator parameter graph node and the processing data graph node based on the second parameter association feature to form a parameter association directed graph network; Acquiring historical laser printing data from the laser printing process database, wherein the historical laser printing data includes historical process parameters, historical printing processing data, and historical indicator performance parameters with parameter association tags; Using the historical laser printing data to complete the training process of the performance parameter prediction model; Inputting the parameter association directed graph network into the trained performance parameter prediction model, and aggregating second-order neighbor node features with the index parameter graph node as the target node through the performance parameter prediction model; A prediction result is output based on the second-order neighbor node characteristics, and the prediction result is converted into a prediction performance parameter corresponding to the initial process parameter.
8. The intelligent controlled laser lateral drawing transfer printing method according to claim 5, characterized in that: The adjusting the initial process parameters to the optimal process parameters by comprehensively analyzing the root mean square errors among the measured performance parameters, the predicted performance parameters and the index performance parameters comprises the following steps: Calculating a first root mean square error between the measured performance parameter and the predicted performance parameter; Determining whether the first root mean square error is less than a preset root mean square error threshold; If the first root mean square error is less than the root mean square error threshold, performing weighted average calculation on the measured performance parameter and the predicted performance parameter to obtain a test performance parameter; Calculating a second root mean square error between the test performance parameter and the index performance parameter; Determining whether the second root mean square error is less than the root mean square error threshold; If the second root mean square error is less than the root mean square error threshold, taking the initial process parameters as the optimal process parameters; If the second root mean square error is greater than or equal to the root mean square error threshold, the initial process parameters are iteratively updated based on the parameter association analysis model and the performance parameter prediction model and according to the second root mean square error until the second root mean square error is less than the root mean square error threshold, and the initial process parameters of the current iteration layer are used as the optimal process parameters.
9. The intelligent controlled laser lateral drawing transfer printing method according to claim 8, characterized in that: The method further comprises the steps of: If the first root mean square error is greater than or equal to the root mean square error threshold, model parameters of the parameter association analysis model and the performance parameter prediction model are iteratively updated based on the first root mean square error until the first root mean square error is less than the root mean square error threshold.
10. An intelligently controlled laser transverse wire drawing transfer printing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the intelligently controlled laser transverse wire drawing transfer printing method as described in any one of claims 1 to 9 is implemented.
Citation Information
Patent Citations
Prediction method of shot peening process parameters based on genetic algorithm optimized BP neural network
CN109508488A
Processing equipment process parameter optimization method and system
CN111597729A
Big data-based printing equipment remote control method and system, and medium
CN114553927A
UV three-dimensional printing optimization method, device and equipment and readable storage medium
CN116766581A
Intelligent control method and system for in-mold label processing equipment
CN117734332A
Cited By
Photoetching positioning paper printing method with image-text suspension dynamic effect
CN120503527A
Plastic manufacturing process parameter optimization method and system based on neural network
CN120633474A
A Neural Network-Based Method and System for Optimizing Plastic Manufacturing Process Parameters
CN120633474B
