A multi-objective collaborative optimization method and system for process parameters of film spraying printing

Abstract

This invention discloses a multi-objective collaborative optimization method and system for film coating printing process parameters, specifically relating to the field of process parameter optimization technology. It addresses the problem of print quality degradation in boundary areas caused by neglecting the process coupling effect between printheads during multi-printhead collaborative operation. By analyzing historical time series and quality data, the method identifies the performance correlation between printhead units, extracts and quantifies coupling influence patterns, constructs a pattern network based on causal analysis, obtains coupling pattern families through community partitioning, evaluates the current task conditions to match relevant pattern families, and integrates their features to generate composite coupling influence patterns as collaborative constraints. A collaborative optimization model is then constructed and solved to obtain a collaborative process parameter combination that can coordinate the operation of multiple printheads. This method mines and utilizes the inherent coupling laws between printheads from data, transforming system-level collaborative requirements into specific optimization constraints, effectively improving the overall print quality during multi-printhead collaborative operation.

Description

Technical Field

[0001] This invention relates to the field of process parameter optimization technology, and in particular to a multi-objective collaborative optimization method and system for film spraying and printing process parameters. Background Technology

[0002] In the field of film coating and printing, especially in the manufacturing of large-area or high-precision functional patterns, multi-head parallel operation or multi-pass scanning and splicing can be used to improve production efficiency. In existing technologies, the optimization of process parameters usually focuses on a single printhead or an independent printing pass. Specifically, by establishing a mapping model between process parameters and printing quality indicators, such as coating thickness, line width, and uniformity, and using optimization algorithms to solve the model, the recommended parameter combination within the independent work unit can be obtained, thereby achieving optimal performance within a single work unit.

[0003] However, existing optimization methods have shortcomings when dealing with multi-unit collaborative operation scenarios. When multiple independently optimized printheads or printing passes are spliced ​​together, the performance differences between units and the physical coupling effects in the unit boundary areas, such as mutual thermal interference and ink fusion, can lead to a decrease in overall output quality. Existing technologies treat the system as a simple superposition of multiple independent units, and their optimization models cannot characterize and optimize the correlation between units. As a result, the local optimal parameter combination may cause global collaborative optimization failures such as uneven quality in the splicing area and sudden performance changes after system integration. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a multi-objective collaborative optimization method and system for film coating and printing process parameters.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A multi-objective collaborative optimization method for membrane coating and printing process parameters includes: S1. Obtain the independent process parameter set, corresponding print quality dataset, and print timing data for each printhead unit; S2. Based on the print quality dataset and print timing data, analyze the contribution ratio of the working timing of each printhead unit to the print quality of the boundary area, and determine the performance correlation between printhead units. S3. Extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and calculate the causal information transmission strength between any two coupling influence modes. S4. Construct a directed weighted network of coupled influence patterns based on the strength of causal information transmission, and obtain multiple families of coupled patterns through community partitioning; S5. Evaluate the matching degree between the current task condition characteristics and each coupling mode family, and integrate the characteristics of multiple related coupling mode families to generate a composite coupling influence mode as a cooperative constraint condition. S6. Using the set of independent process parameters as optimization variables, the global printing quality index as the optimization objective, and constructing a multi-nozzle collaborative optimization model with collaborative constraints, the combination of multi-nozzle collaborative process parameters that satisfy the collaborative constraints is obtained.

[0006] Furthermore, S1 includes: Obtain the independent set of process parameters for each printhead unit in historical printing tasks; Obtain the print quality dataset corresponding to each historical print task. The print quality dataset contains print quality measurements of the boundary areas. Obtain the printing timing data of each printhead unit when executing each historical printing task. The printing timing data includes the jet start and stop time sequence of each printhead unit.

[0007] Furthermore, S2 includes: Based on the print quality dataset and print timing data, the print timing data of each printhead unit is time-aligned with the print quality measurement values ​​of the boundary region in the print quality dataset; A linear model is constructed using the printing timing data of each printhead unit after timing alignment and the printing quality measurement values ​​of the boundary area; the proportion of variance explained by the main effect of the working timing of each printhead unit to the total explained variance of the model is calculated by variance decomposition, which is taken as the contribution ratio of each printhead unit. Based on the contribution ratio of each nozzle unit, determine whether there is a performance correlation between any two nozzle units and the strength of the correlation.

[0008] Furthermore, S3 includes: Based on performance correlation, the process parameter combinations corresponding to adjacent printhead units that have a coupled impact on the printing quality of the boundary area are identified; The identified process parameter combinations and the boundary area printing quality measurement values ​​are parametrically modeled to obtain the parametric relationship with process parameters as independent variables and boundary area printing quality measurement values ​​as dependent variables as the coupling influence mode. For any two coupled influence modes, the causal information transmission strength from one coupled influence mode to another is calculated based on the transfer entropy algorithm.

[0009] Furthermore, the calculation of causal information transmission strength based on the transfer entropy algorithm includes: treating the data sequences corresponding to the two coupled influence modes as two random processes; calculating the conditional mutual information from the past state of the source random process to the current state of the target random process to obtain the transfer entropy value, which is used as the causal information transmission strength from the source coupled influence mode to the target coupled influence mode.

[0010] Furthermore, S4 includes: Each coupled influence pattern is treated as a network node, and the causal information transmission strength between any two coupled influence patterns is used as the weight of the directed edge between the corresponding nodes to construct a directed weighted network. The Louvain community detection algorithm is used to perform community partitioning on the constructed directed weighted network. The set of coupling influence patterns contained in each divided community is defined as a family of coupling patterns.

[0011] Furthermore, the community partitioning using the Louvain community discovery algorithm includes: iteratively performing modularity optimization and community merging, continuously assigning nodes to communities that maximize network modularity, and merging communities into new nodes until modularity is stable, and finally defining the set of coupling influence patterns within each stable community as a coupling pattern family.

[0012] Furthermore, S5 includes: Based on the working characteristics of the current task and the characteristics of the coupling influence modes in each coupling mode family, calculate the matching degree between the current task and each coupling mode family. Based on the matching degree, select the families of coupling patterns with the highest matching degree as the relevant coupling pattern families; The features of each selected family of related coupling modes are weighted and fused according to their corresponding matching degree to generate a composite coupling influence mode. The generated complex coupling influence pattern is used as a cooperative constraint condition.

[0013] Furthermore, S6 includes: Using the independent process parameters of each nozzle unit in the independent process parameter set as optimization variables, maximizing the global printing quality index as the optimization objective, and the parameter relationship characterized by the complex coupling influence mode as the inequality constraint, a constrained multi-objective optimization model is constructed. A multi-objective evolutionary algorithm is used to solve the constructed constrained multi-objective optimization model. A set of solutions is selected from the obtained Pareto optimal solution set, and the independent process parameter values ​​of each nozzle unit corresponding to these solutions are used as the multi-nozzle collaborative process parameter combination that satisfies the collaborative constraint conditions.

[0014] On the other hand, the present invention provides a multi-objective collaborative optimization system for film coating and printing process parameters, comprising: The data acquisition module is used to acquire the independent set of process parameters, the corresponding print quality dataset, and the print timing data for each printhead unit. The relationship judgment module is used to analyze the contribution ratio of the working time sequence of each printhead unit to the printing quality of the boundary area based on the print quality dataset and print timing data, and to determine the performance correlation between printhead units. The intensity calculation module is used to extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and to calculate the causal information transmission intensity between any two coupling influence modes. The community partitioning module is used to construct a directed weighted network of coupled influence patterns based on the strength of causal information transmission, and obtains multiple families of coupled patterns through community partitioning; The condition generation module is used to evaluate the matching degree between the current task condition characteristics and each coupled mode family, and to integrate the features of multiple related coupled mode families to generate a composite coupled influence mode as a collaborative constraint condition. The combined solution module is used to construct a multi-nozzle collaborative optimization model with independent process parameter sets as optimization variables, global printing quality index as optimization objective, and collaborative constraints, and solve for the multi-nozzle collaborative process parameter combination that satisfies the collaborative constraints.

[0015] The beneficial effects of this invention are: 1. Through a data-driven approach, the inherent process coupling relationships in multi-printer collaborative operations are systematically identified and quantified, and transformed into collaborative constraints in a global optimization model. This allows the optimization process to move beyond treating the system as a simple collection of isolated units, actively modeling and utilizing the interactions between units. This effectively overcomes the problems of print quality degradation in boundary areas and collaborative optimization failure caused by ignoring coupling effects. A set of coordinated multi-printer collaborative process parameters at the system level is generated, ensuring that locally optimized parameters, after integration, not only do not cause performance conflicts, but also improve the consistency of print quality in splicing areas through collaborative cooperation, achieving a leap from local optimization to global collaboration.

[0016] 2. By constructing a complete analysis chain from time series analysis and correlation mining to pattern extraction and network partitioning, the interaction patterns between printheads implicit in complex data are made explicit and structured into computable constraint knowledge. This enables the collaborative optimization model to have decision memory based on historical experience. While pursuing traditional global quality indicators, its optimization objective must also meet the constraints derived from the data that reflect the real physical coupling laws. The final combination of process parameters not only mathematically optimizes the overall quality indicators, but also ensures the matching and synergy of the working states of multiple printheads in terms of physical mechanisms. This significantly improves the scientificity and reliability of process parameter setting in complex film printing tasks and reduces the trial and error costs and quality risks caused by parameter mismatch. Attached Figure Description

[0017] Figure 1This is a flowchart of a multi-objective collaborative optimization method for membrane material spraying and printing process parameters according to the present invention; Figure 2 This is a schematic diagram of the structure of a multi-objective collaborative optimization system for film coating and printing process parameters according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention presents a multi-objective collaborative optimization method for film coating and printing process parameters, comprising: S1. Obtain the independent process parameter set, corresponding print quality dataset, and print timing data for each printhead unit; S2. Based on the print quality dataset and print timing data, analyze the contribution ratio of the working timing of each printhead unit to the print quality of the boundary area, and determine the performance correlation between printhead units. S3. Extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and calculate the causal information transmission strength between any two coupling influence modes. S4. Construct a directed weighted network of coupled influence patterns based on the strength of causal information transmission, and obtain multiple families of coupled patterns through community partitioning; S5. Evaluate the matching degree between the current task condition characteristics and each coupling mode family, and integrate the characteristics of multiple related coupling mode families to generate a composite coupling influence mode as a cooperative constraint condition. S6. Using the set of independent process parameters as optimization variables, the global printing quality index as the optimization objective, and constructing a multi-nozzle collaborative optimization model with collaborative constraints, the combination of multi-nozzle collaborative process parameters that satisfy the collaborative constraints is obtained.

[0020] S1. Obtain the independent process parameter set, corresponding print quality dataset, and print timing data for each printhead unit. Specifically, this is implemented as follows: First, the independent set of process parameters for each printhead unit in historical printing tasks is obtained. Historical printing tasks refer to previously completed and archived film coating printing operations. For each historical printing task, the specific process parameter values ​​used by each printhead unit in that task are recorded. These parameters constitute the independent set of process parameters for that printhead unit in that task. The independent set of process parameters for a printhead unit typically includes jet pressure, jet frequency, jet pulse width, ink temperature, and the distance between the printhead and the film substrate. Jet pressure refers to the pressure of the gas or liquid driving the ink out of the nozzle, measured in megapascals (MPA). In historical tasks, this parameter may range from, for example, 0.1 MPA to 0.5 MPA. Jet frequency refers to the number of jet cycles the printhead completes per unit time, measured in kilohertz (KHz). In historical tasks, this parameter may range from, for example, 1 KHz to 10 KHz. Jet pulse width refers to the duration for which the drive voltage or drive signal remains active during each jet, measured in microseconds (µs). Ink temperature refers to the actual temperature of the ink supplied to the printhead, measured in degrees Celsius. The distance between the printhead and the substrate refers to the vertical distance from the printhead tip to the surface of the substrate being printed, measured in millimeters. This historical data is extracted from the historical process database of the control system of the inkjet printing equipment. This database stores data in tabular form, with each record containing at least four fields: historical print job number, printhead unit number, process parameter name, and process parameter value. These fields allow all independent process parameters used by each printhead unit in each job to be summarized into its independent process parameter set.

[0021] Secondly, a print quality dataset corresponding to each historical printing task is obtained. This dataset explicitly includes print quality measurements of the boundary regions. After each historical printing task is completed, the printed film samples need to undergo specialized quality inspection to form the print quality dataset. The boundary region specifically refers to the area where the printing passes of two or more printhead units are spatially adjacent or overlapping when using multi-head collaborative operation. The physical location and extent of this region can be predefined based on the printhead layout and scanning path. Print quality measurements are the results obtained by quantitatively inspecting these predefined boundary regions. Specifically, this includes the average thickness, thickness uniformity, sheet resistance, and straightness of the pattern edges within the boundary region. The average thickness is measured using a contact thickness gauge or laser thickness gauge at multiple points within the boundary region according to a preset measurement point matrix. The thickness values ​​at all measurement points are then arithmetically averaged, with the unit being micrometers. Thickness uniformity is quantified by calculating the standard deviation of the same set of thickness measurements, also in micrometers. Sheet resistance is measured using a four-probe tester at the center of the boundary region, with the unit being ohms per square meter. The straightness of the pattern edges is determined by acquiring images of the boundary region using an optical microscope or line scan camera, then extracting the pattern edges using image processing algorithms and calculating their average deviation from an ideal straight line, measured in micrometers. Each historical printing task corresponds to a print quality dataset, a structured collection where each record contains at least a historical printing task number, a boundary region identifier number, and one or more print quality measurements. This data is exported and stored from independent quality inspection instruments or integrated online inspection systems and is strictly correlated with the corresponding historical printing task number.

[0022] Finally, the printing timing data of each printhead unit during the execution of each historical print job is obtained. This timing data includes the sequence of jet start and stop times for each printhead unit. During the execution of a historical print job, the device's motion controller or dedicated data acquisition card records the state of the jet enable signal for each printhead unit at a fixed sampling period. The jet start and stop time sequence is extracted from these continuous signal state records. It consists of a series of paired timestamps, each pair corresponding to the start and end times of a jet action. The time value is a relative time calculated from the start time of the job, and its unit can be milliseconds or microseconds, depending on the clock accuracy of the control system, for example, it can reach an accuracy of 1 millisecond. For a given printhead unit, the complete printing timing data in a specific historical print job includes the start and stop time pairs of all jet events of that printhead throughout the entire job cycle, arranged in chronological order. This timing data is exported in time-series format from the controller's operation log file or real-time database. The exported data file or record explicitly contains the historical print job number, printhead unit number, event type, and the timestamp when the event occurred. Through the above three specific and coherent data acquisition actions, a complete data foundation is obtained on which all subsequent analysis steps depend. These data can be accurately and uniquely matched through the common historical printing task number and nozzle unit number as key correlation fields, thereby ensuring that a clear, traceable logical relationship and data integrity are established from the data acquisition stage.

[0023] S2. Based on the print quality dataset and print timing data, analyze the contribution ratio of each printhead unit's working timing to the print quality of the boundary area, and determine the performance correlation between printhead units. The specific implementation is as follows: First, based on the print quality dataset and print timing data, the print timing data of each printhead unit is time-aligned with the print quality measurements of the boundary region in the print quality dataset. The purpose of this time alignment is to establish a temporal correspondence between the ejection action of each printhead unit and the final print quality measurement observed in the boundary region on a unified time reference. Specifically, the start time of each independent historical print task is used as the time zero point. The sequence of ejection start and stop times for each printhead unit recorded in the print timing data is an offset relative to the start time of this task, typically in milliseconds. The print quality measurements of the boundary region in the print quality dataset are obtained after the task is completed. To correlate them with the time of the printing process, calculations are needed based on the scanning motion speed of the printhead and the spatial position of the boundary region. The scanning motion speed is a known process parameter recorded by the printing equipment during the execution of the historical print task, for example, a value of 100 mm / s. The spatial position of the boundary region refers to its start and end coordinates on the film substrate, which are determined by the print path planning file. The length of the boundary region along the scanning direction can be obtained from the difference between the starting and ending coordinates; for example, this length is 5 mm. The approximate printing time window length corresponding to the formation of the film layer in this boundary region can then be calculated by dividing the boundary region length by the scanning speed, i.e., 5 / 100 = 0.05 seconds. The starting point of this time window needs to be determined by combining the precise moment when the printhead moves to the starting coordinate of the boundary region, which can be obtained from the device's motion control log. By comparing the jetting event time interval of each printhead unit with the calculated formation time window of each boundary region, if the jetting time interval of a printhead unit overlaps with the formation time window of a certain boundary region, it is considered that the working timing of that printhead unit participated in the formation of that boundary region, thus completing the timing alignment between the printhead timing and the boundary quality measurement value. This process establishes an accurate data foundation for subsequent analysis, ensuring that the printing quality measurement value of each boundary region can be accurately correlated with the timing activities of all printhead units that have performed jetting actions within their formation time window.

[0024] Secondly, a linear model is constructed using the timing data of each printhead unit after timing alignment and the print quality measurements of the boundary region. Variance decomposition is then used to calculate the proportion of variance explained by the main effects of each printhead unit's operating timing to the total explained variance of the model, which is then used as the contribution ratio of each printhead unit. When constructing the linear model, the print quality measurement of each timing-aligned boundary region is used as the dependent variable of the model, with units consistent with its measurement unit, such as thickness in micrometers. For each boundary region corresponding to this dependent variable, the model's input variable, i.e., the independent variable, is the quantified timing characteristics of all printhead units that have performed spraying actions within the time window of that boundary region, determined according to the timing alignment relationship. These quantified timing characteristics are extracted from the original spray start-stop time sequence. For a given printhead unit within a given boundary region formation time window, multiple features can be extracted, such as the cumulative spraying time of the printhead within this time window (the sum of the durations of all spraying events, in milliseconds); or the number of sprays by the printhead within this time window; or the proportion of spraying time by the printhead within this time window (the cumulative spraying time divided by the total length of the time window). Typically, one of the most representative features is selected as the temporal characteristic value of the printhead unit in this sample. The temporal characteristic values ​​of all relevant printhead units are used as independent variables, together with the dependent variable of the boundary area print quality measurement, to form a training data sample for a multiple linear regression model. By fitting a large number of such samples from multiple historical printing tasks and estimating the model parameters using the least squares method, a linear model reflecting the statistical relationship between the printhead's working temporal characteristics and the boundary print quality can be obtained. After obtaining this linear model, variance decomposition is used to analyze the independent explanatory power of each independent variable for the variation of the dependent variable.

[0025] Variance decomposition breaks down the total variance explained by the model into the main effect variances of each independent variable and the interaction effect variances between the independent variables. The variance explained by the main effect of the operating time of a printhead unit refers to the amount of variance of the dependent variable that the fitted simple linear regression model can explain when the model includes only the operating time characteristics of that printhead unit as the sole independent variable. The contribution ratio of a printhead unit is calculated by dividing the variance explained by its operating time main effect by the sum of the variances explained by the main effects of all relevant printhead units in the model. For example, if the operating time characteristics of printhead unit A explain 8 units of variance, printhead unit B explains 12 units, printhead unit C explains 20 units, and the total variance explained by the main effects of all printhead units is 40 units, then the contribution ratio of printhead unit A is 8 / 40 = 0.2. This contribution ratio quantitatively characterizes the degree of independent influence, statistically speaking, of the variation in the operating time of the printhead unit on the variation of the print quality measurement value in the boundary area.

[0026] Finally, based on the contribution ratio of each printhead unit, the existence and strength of a performance correlation between any two printhead units are determined. Here, performance correlation specifically refers to the statistical correlation or mutual influence between the contribution patterns of the working sequences of two printhead units on the printing quality of the boundary area. The determination method is based on the following logic and steps: First, for each printhead unit, its contribution ratio calculated using the linear model and variance decomposition in all historical task samples is collected, forming a contribution ratio sequence for that printhead unit. This sequence is a list of values, with each value corresponding to a historical task sample. For any two printhead units to be judged, such as printhead unit X and printhead unit Y, their respective contribution ratio sequences are extracted. The statistical correlation coefficient between these two sequences is calculated, such as the Pearson correlation coefficient. The calculation of the Pearson correlation coefficient requires two numerical sequences as input, and its output is a value between -1 and +1. The absolute value of this value can be used to quantify the strength of the linear correlation between the two sequences. To make a binary classification judgment of the existence or non-existence of a performance correlation, a performance correlation judgment threshold needs to be set. The threshold for determining the performance correlation is not a fixed value, but is determined statistically based on the distribution characteristics of historical data. Specifically, for all nozzle units involved in the analysis, the absolute values ​​of the Pearson correlation coefficients of their pairwise contribution ratio sequences are calculated, resulting in a set of absolute correlation coefficient values. The upper quartile of this set is taken as the performance correlation threshold. For example, if the upper quartile of the set of absolute correlation coefficient values ​​for all nozzle units is 0.6, then 0.6 is set as the performance correlation threshold. If the absolute value of the correlation coefficient between the contribution ratio sequences of nozzle units X and Y is greater than or equal to 0.6, a performance correlation is determined to exist between these two nozzle units; if it is less than 0.6, no significant performance correlation is determined. The strength of the correlation is continuously represented by the absolute value of the correlation coefficient; the larger the absolute value, the stronger the correlation. Simultaneously, the sign of the correlation coefficient indicates the directionality of the correlation: a positive sign indicates that the contribution ratios of the two nozzle units tend to change in the same direction, while a negative sign indicates that they tend to change in opposite directions.

[0027] S3. Extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and calculate the causal information transmission strength between any two coupling influence modes. The specific implementation is as follows: First, based on performance correlation, the process parameter combinations corresponding to adjacent printhead units that have a coupled impact on the print quality of the boundary area are identified. Adjacent printhead units here refer to two printhead units in the physical layout of the printing equipment whose nozzle array center distance is less than a preset distance threshold. This preset distance threshold is determined based on the minimum interval of the equipment's mechanical structure, for example, 20 mm. Alternatively, it refers to printhead units configured to sequentially print the same film material area in the timing logic of the printing task. The performance correlation output in step S2 clearly identifies whether there is a correlation between any two printhead units and the strength of that correlation. The specific criterion for identifying the coupling effect is: when the print quality measurement value of a boundary area is determined to be significantly affected by the working timing of two or more printhead units with performance correlation, this group of printhead units is marked as having a coupled effect on the boundary area. Next, to obtain the process parameter combinations corresponding to these printhead units, it is necessary to trace back all relevant historical printing tasks. Specifically, for each labeled nozzle unit group with coupled influence, such as nozzle unit A and nozzle unit B, all independent process parameter values ​​for nozzle unit A and nozzle unit B in each relevant historical task are extracted from their respective independent process parameter sets. These values ​​are then arranged and merged in a fixed order to form a complete process parameter vector representing this nozzle unit combination in that task. This vector is a process parameter combination instance. For example, the process parameters for nozzle unit A might include an injection pressure of 0.25 MPa and an injection frequency of 5 kHz, while the process parameters for nozzle unit B might include an injection pressure of 0.28 MPa and an injection frequency of 4.8 kHz. By summarizing all process parameter combination instances from relevant historical tasks, a process parameter combination dataset for this nozzle unit group is constructed for subsequent modeling.

[0028] Secondly, the identified combinations of process parameters and the printed quality measurements of the boundary areas are parametrically modeled to obtain a parametric relationship as the coupling influence mode, with process parameters as independent variables and the printed quality measurements of the boundary areas as dependent variables. The purpose of parametric modeling is to establish a quantitative mapping relationship between the numerical values ​​of the process parameter combinations and the printed quality measurements of the boundary areas. The input data for modeling is paired, with each pair including an instance of a process parameter combination and a corresponding printed quality measurement of the boundary area. Taking printhead units A and B as an example, the input of a data sample is a vector of process parameters, and the output is the average printed quality measurement of one or more boundary areas jointly affected by A and B in this historical task, for example, an average thickness of 12.5 micrometers. Multiple regression analysis methods are used for modeling, such as multinomial regression or support vector regression. Before modeling, all process parameter values ​​need to be standardized to eliminate the influence of different parameter dimensions. Standardization methods include subtracting the historical mean of each parameter value and then dividing by the historical standard deviation. The model training process is achieved by minimizing the error between the predicted value and the actual measured value, for example, minimizing the mean square error. By training and cross-validating on multiple sets of historical data, the structural form and internal coefficients of the model are finally determined, resulting in a stable and reproducible parameterized relationship. This determined parameterized relationship is defined as a coupling effect mode. Each coupling effect mode corresponds to a specific group of adjacent nozzle units with coupled effects. The input of its function expression is all the process parameters of that nozzle unit group, and the output is a predicted value of the printing quality of the boundary area caused by their combined effect.

[0029] Finally, for any two coupled influence modes, the causal information transmission strength from one coupled influence mode to another is calculated based on the transfer entropy algorithm. Before the calculation, it is necessary to construct a unique residual data sequence for each coupled influence mode. Specifically, the parameterized function of the coupled influence mode is applied to all relevant historical printing tasks. For each historical task, the actual process parameter values ​​of the corresponding printhead unit group in the task are substituted into the function to calculate the predicted value of the printing quality of a boundary area; at the same time, the actual boundary area printing quality measurement value is obtained from the printing quality dataset of the task; the difference between the predicted value and the actual value is calculated, i.e., the residual. Arranging all historical tasks in chronological order, the corresponding residual value sequence constitutes the residual data sequence of the coupled influence mode. This sequence reflects the error information that the mode could not fully explain in history and that varied with the task. Next, the causal information transmission strength between the two coupled influence modes, denoted as mode A and mode B, is calculated. The residual sequence of mode A is regarded as a stochastic process X, and the residual sequence of mode B is regarded as a stochastic process Y. The transfer entropy algorithm calculates the conditional mutual information from the past states of random process X to the current state of random process Y. The implementation process includes the following core steps: First, a time delay parameter L is set. This parameter defines the length of the time window for past states. The value of L is determined based on the temporal correlation of the specific task sequence. For example, it can be determined by calculating the autocorrelation of the sequence. L can be set to 2, indicating that the states of the past two historical tasks are being examined. Second, the continuous residual values ​​are discretized and preprocessed, dividing them into a finite number of state intervals. The number of state intervals K is a preset parameter. For example, the residual range is divided into 5 equal intervals, i.e., K equals 5. Each residual value is mapped to a corresponding interval number. Third, based on the discretized state sequence, the joint probability of the simultaneous occurrence of the state combinations of random process X at the past L time points, the state of random process Y at the current time point, and the state combinations of random process Y itself at the past L time points is calculated. Simultaneously, the conditional probability of the simultaneous occurrence of the state of random process Y at the current time point and its own state combinations at the past L time points is also calculated. Fourth, according to the definition of conditional mutual information, a specific transfer entropy value is calculated using the probabilities obtained above. This value is defined as the causal information transmission strength from source coupling influence mode A to target coupling influence mode B. To determine whether the causal information transmission strength is significant, a causal significance threshold needs to be set. This threshold is obtained by first calculating the transmission entropy values ​​between each pair of all coupling influence modes to be evaluated, forming a set of transmission entropy values; then calculating the average of this set of transmission entropy values; and finally multiplying this average by a coefficient greater than 1, such as a coefficient of 2. The resulting product is the causal significance threshold.If the transfer entropy value calculated from mode A to mode B is greater than or equal to the causal significance judgment threshold, then a significant causal influence is determined to exist between mode A and mode B; if the transfer entropy value is less than the causal significance judgment threshold, then no significant causal influence is determined to exist. By systematically calculating the transfer entropy values ​​between all pairs of coupled influence modes and comparing them with the causal significance judgment threshold, a complete and qualitative causal influence relationship network can be obtained. At the same time, the transfer entropy value itself serves as a quantitative strength, providing accurate weight data for network construction in step S4.

[0030] S4. Construct a directed weighted network of coupled influence patterns based on the strength of causal information transmission, and obtain multiple families of coupled patterns through community partitioning. The specific implementation is as follows: First, each coupled influence mode is treated as a network node, and the causal information transmission strength between any two coupled influence modes is used as the weight of the directed edge between the corresponding nodes, thus constructing a directed weighted network. The specific input to the construction process is the causal information transmission strength data between all pairs of coupled influence modes calculated in step S3. This data constitutes a causal information transmission strength matrix, where each row and column index corresponds to a unique coupled influence mode. Each element in the matrix represents the causal information transmission strength from the row index mode to the column index mode; this strength value is a dimensionless non-negative real number, and a larger value indicates a stronger causal influence. During network construction, each independent coupled influence mode is mapped to a network node, and the node identifier is typically the number of the adjacent nozzle unit pair corresponding to that mode. For any two different nodes, i.e., two different coupled influence modes, it is necessary to determine whether a directed edge is established between them and the weight of that edge based on the causal information transmission strength matrix. Here, a threshold for edge weight generation needs to be set to filter out weak or insignificant causal relationships. The method for setting the edge weight generation threshold is as follows: Extract all off-diagonal elements from the causal information transmission strength matrix, i.e., the strength values ​​between all different pattern pairs, forming a set of strength values; calculate the median of this set of strength values; and use this median value as the edge weight generation threshold. Only when a causal information transmission strength value is greater than or equal to this edge weight generation threshold is a directed edge established between the corresponding two nodes, with the edge pointing from the source pattern node to the target pattern node, and the edge weight directly assigned to the causal information transmission strength value. If the strength value is less than this edge weight generation threshold, no edge is established, indicating that there is no significant direct causal influence between the two patterns under the current threshold. If all strength values ​​are lower than this edge weight generation threshold, each node will become an isolated node without any edges. By traversing all pattern pairs and establishing nodes and edges according to this rule, a directed weighted network consisting of a set of nodes and a set of directed edges is finally formed. This network is stored in computer memory as a graph data structure, which can intuitively represent the topological structure of causal interactions between different coupled influence patterns.

[0031] Secondly, the Louvain community detection algorithm is used to partition the constructed directed weighted network into communities. The purpose of this step is to identify subsets of nodes with close connections and frequent internal causal interactions, grouping them into the same community. The Louvain community detection algorithm is an iterative algorithm aimed at optimizing network modularity. Modularity is an indicator of the quality of community partitioning; a higher value indicates closer connections within a community and sparser connections between communities. The algorithm's execution process involves two iterative phases. The first phase is local optimization of modularity. Initially, each node in the network is considered an independent community. Then, each node in the network is traversed, and for the current node, the communities to which all its neighboring nodes belong are examined. Attempts are made to move the current node to each neighboring community, and the network modularity gain resulting from each virtual move is calculated. The calculation of the modularity gain requires consideration of the current network's edge weights and community structure. The principle is to compare the increased edge weights within the new community after a node's movement with the potential decrease in inter-community edge weights due to the movement, calculating a gain value through a defined mathematical relationship. This gain value can be positive or negative. If one or more adjacent communities can generate a positive modularity gain, the node is moved to the community that maximizes the modularity gain. If no moves produce a positive modularity gain, the node remains in its original community. This process is repeated, traversing all nodes, until no move can further improve the overall network modularity, at which point a locally optimal community partitioning is achieved. The second stage is community merging, where each community obtained in the first stage is considered a new supernode, and all nodes originally belonging to the same community are shrunk to this supernode. The edge weights between the new supernodes are the sum of the weights of all edges between nodes belonging to two different communities in the original network. This results in a new, smaller weighted network. Then, the modularity optimization process from the first stage is applied to this new network, and node moves and community partitioning are performed again. These two stages—local modularity optimization and community merging—constitute a complete iteration. The algorithm repeatedly executes such iterations, constructing and optimizing a new network based on the results of the previous round. The algorithm terminates when the network's modularity index no longer increases, i.e., the difference between the network modularity values ​​calculated in two adjacent iterations is less than a preset algorithm stopping threshold. The algorithm uses a very small positive threshold, such as 0.0001, to determine whether the optimization process has converged. To prevent infinite loops, a maximum number of iterations, such as 1000, is set; the algorithm is forced to stop when this number is reached. Once the stopping condition is met, the community partitioning result obtained in the first stage of the last iteration is the final community partitioning scheme.

[0032] Finally, the set of coupled influence modes contained within each divided community is defined as a coupled mode family. Based on the final community partitioning result output by the Louvain algorithm, each community contains one or more network nodes, and each node corresponds to a coupled influence mode defined in step S3. The coupled influence modes corresponding to all nodes within a community are grouped together to form a coupled mode family. Modes within each coupled mode family indicate that they are closely connected in the causal influence network, exhibiting strong direct or indirect causal interactions, potentially reflecting a specific, inherently unified process parameter coupling mechanism. However, the causal connections between coupled influence modes from different coupled mode families are relatively weaker. For use in subsequent steps, each coupled mode family is assigned a unique family identifier, and the specific information of all coupled influence modes it contains is recorded, including the adjacent nozzle unit pairs corresponding to each mode and their parameterized models. Through this step, a large number of complex coupled influence modes are effectively classified into several coupled mode families, achieving dimensionality reduction and structured organization of the modes, laying the foundation for subsequent targeted matching and fusion of mode features.

[0033] S5. Evaluate the matching degree between the current task condition characteristics and each coupled mode family, and fuse the features of multiple related coupled mode families to generate a composite coupled influence mode as a cooperative constraint condition. The specific implementation is as follows: First, based on the operational characteristics of the current task and the characteristics of the coupling influence modes in each coupling mode family, the matching degree between the current task and each coupling mode family is calculated. The operational characteristics of the current task refer to a set of feature vectors describing the external conditions and requirements of the new printing task to be executed. This feature vector includes at least three elements: the type of film material to be printed, the width of the film substrate, and the expected average thickness of the overall printing task. The value of the film material type comes from the production order of this new printing task, such as polyethylene terephthalate or polyimide. The unit of the film substrate width is millimeters, for example, 1000 millimeters. The unit of the expected average thickness is micrometers, for example, 10 micrometers. The characteristics of the coupling influence modes in each coupling mode family refer to an abstract feature vector characterizing the commonalities of all coupling influence modes within that family. The feature vector extraction method is as follows: For a given family of coupled patterns, collect all coupled influence patterns it contains; analyze the coefficients of the process parameter independent variables in the parameterized model corresponding to each pattern, such as the coefficients of injection pressure and injection frequency in a pattern function expression; calculate the average value of all patterns within the family across these coefficient dimensions, thus forming a mean coefficient vector. Each dimension of the mean coefficient vector corresponds to the influence intensity of a process parameter on the boundary printing quality, and this intensity is a dimensionless numerical value; simultaneously, statistically analyze the common physical layout types of the nozzle unit pairs involved in all patterns within the family, such as "adjacent left and right" or "sequential front and back," and encode them as a categorical feature. Combining the mean coefficient vector with the layout categorical feature constitutes the feature vector of the coupled pattern family. The matching degree is calculated by comparing the similarity between the feature vector of the current task condition and the feature vector of a coupled pattern family. Since the feature vector contains numerical and categorical dimensions, the similarity calculation needs to be processed separately. For the numerical part, the cosine similarity between two vectors is calculated. Specifically, the corresponding numerical dimensions of the two vectors are multiplied, summed, and then divided by the product of the magnitudes of the two vectors (i.e., the square root of the sum of the squares of their elements). This yields a numerical similarity score between -1 and +1. For the categorical part, a consistency check is performed. If the membrane material type matches the typical membrane material type required by the nozzle layout, a categorical similarity score of 1 is assigned; otherwise, it is 0. Finally, a weighted average is calculated between the numerical and categorical similarity scores. This involves multiplying the numerical similarity score by a numerical weight and the categorical similarity score by a categorical weight, with the sum of the two weights being 1. For example, if the numerical weight is 0.7 and the categorical weight is 0.3, the sum yields a final matching score between 0 and 1. This value represents the degree of adaptation between the current task condition and the coupling mechanism represented by this family of coupling modes.

[0034] Secondly, based on the matching degree ranking, the highest-ranking coupled pattern families are selected as relevant coupled pattern families. After calculating the matching degree for all coupled pattern families, a list of matching degree values ​​is obtained. The matching degree values ​​in this list are then sorted in descending order. The selection of relevant coupled pattern families requires a minimum matching degree threshold and a selection quantity threshold. The minimum matching degree threshold is set based on historical experience, for example, 0.5, to exclude coupled pattern families with excessively low matching degrees. The selection quantity threshold is set as follows: First, the number M of coupled pattern families with matching degrees greater than the minimum matching degree threshold is counted; then, the selection quantity N is calculated, which is M multiplied by a preset proportional coefficient. This proportional coefficient is set according to the requirements for the breadth of pattern fusion in actual applications, for example, 0.3; if the calculated N is not an integer, it is rounded up; if N is less than 1, it is set to 1. According to this rule, the coupled pattern families ranked 1st to Nth in the matching degree ranking list are selected as relevant coupled pattern families. If no coupled pattern family has a matching degree greater than the minimum matching degree threshold, then only the coupled pattern family ranked first in matching degree is selected as the relevant coupled pattern family, and N equals 1.

[0035] Next, the features of the selected related coupled pattern families are weighted and fused according to their corresponding matching degrees to generate a composite coupled influence pattern. Weighted fusion refers to summing the feature vectors of each related coupled pattern family using their matching degree values ​​as weights, thus generating a new, comprehensive feature vector. The specific operation consists of two steps. The first step is weight normalization: calculating the sum S of the original matching degree values ​​of all related coupled pattern families; for each related coupled pattern family, its normalized weight is equal to its original matching degree value divided by S. The second step is weighted summation of the feature vectors: for each dimension of the feature vector, calculating the weighted average of that dimension's features across all related coupled pattern families. The weighted average is calculated by multiplying the feature value of each related coupled pattern family in that dimension by the normalized weight of that family, and then summing the products of all families. For example, regarding the dimension in the feature vector that represents the injection pressure coefficient, suppose there are two related coupled mode families: group A has a value of 0.8 for this dimension with a normalized weight of 0.6; group B has a value of 0.5 for this dimension with a normalized weight of 0.4. Then, the value of this dimension in the composite coupled influence mode is equal to 0.8 × 0.6 + 0.5 × 0.4 = 0.68. This calculation is performed on all dimensions one by one, and the resulting complete new feature vector represents the feature of the composite coupled influence mode. The dimensional structure of this feature vector of the composite coupled influence mode is the same as that of the feature vector of a single coupled mode family.

[0036] Finally, the generated composite coupling influence pattern is used as a collaborative constraint. The feature vector of the above composite coupling influence pattern is converted into a mathematical constraint form that can be used in subsequent optimization steps. This conversion process is as follows: the mean coefficient part of the feature vector of the composite coupling influence pattern defines a linear relationship between process parameters and the predicted print quality value. This linear relationship is expressed as an inequality constraint. Specifically, assuming that the coefficient of the injection pressure of nozzle unit A is C1, the coefficient of the injection pressure of nozzle unit B is C2, and the coefficient of the boundary thickness deviation is C3 in the composite feature, a linear inequality constraint can be constructed in the form: C1×P_A + C2×P_B + C3×ΔT≤δ. Where, P_A and P_B represent the injection pressure of nozzle units A and B respectively, which are optimization variables; ΔT represents the deviation between the predicted boundary thickness and the target thickness; δ is a constraint tolerance threshold. The method for setting the constraint tolerance threshold δ is as follows: Based on the actual boundary thickness deviations corresponding to all sample tasks belonging to the relevant coupling mode family in historical data, calculate the standard deviation σ of these deviation values; then, set δ to k×σ, where k is a coefficient greater than 1 used to control the degree of constraint leniency, for example, k is set to 2. In this way, the complex coupling influence mode is concretized into a clear linear inequality constraint. This constraint will be used as the boundary condition of the optimization model in step S6 to limit the search space of process parameter combinations.

[0037] S6. Using the independent set of process parameters as optimization variables, the global printing quality index as the optimization objective, and constructing a multi-nozzle collaborative optimization model with collaborative constraints, the combination of multi-nozzle collaborative process parameters that satisfies the collaborative constraints is obtained. The specific implementation is as follows: First, taking the independent process parameters of each nozzle unit in the set of independent process parameters as optimization variables, maximizing the global printing quality index as the optimization objective, and using the parameter relationships characterized by the composite coupling influence mode as inequality constraints, a multi-objective optimization model with constraints is constructed. The construction of the model is specifically as follows: Each independent process parameter of all nozzle units involved in the current printing task defined in step S1 is taken as an independent optimization variable, and these optimization variables together form a parameter vector to be optimized. Each optimization variable has its clear physical value range, which is determined based on the hardware performance and safety specifications of the nozzle unit. For example, the lower limit of the injection pressure variable of a nozzle unit is set to 0.1 MPa, and the upper limit is set to 0.5 MPa. The optimization objective, the global printing quality index, is a function that synthesizes multiple global quality evaluation dimensions into a single scalar. Its input is the optimization variable vector, and the output is a numerical value representing the quality of the print. The larger this numerical value, the better the quality. The construction method of this function is: Select two key global quality dimensions, such as the uniformity of the whole surface density U and the proximity of the average thickness to the target value C; respectively set normalization functions for these two dimensions through historical data or expert experience to map their original measured values to scores between 0 and 1; then assign weights w1 and w2 to these two scores, and the sum of the weights is 1. The weight values reflect the emphasis on different quality dimensions. For example, when more importance is attached to uniformity, w1 can be set to 0.7 and w2 to 0.3; finally, the calculation method of the global printing quality index Q is: Q = w1×U + w2×C. The constraint conditions of the model directly adopt the linear inequality constraints transformed from the composite coupling influence mode output in step S5. This constraint is the specific mathematical expression reflecting the coupling relationship between parameters generated in step S5 based on the fusion features, and its form is a linear inequality, such as: C1×P A + C2×P B + C3×ΔT ≤ δ. Among them, the coefficients C1, C2, C3 and the constraint tolerance threshold δ are all known quantities output in step S5, and the variables P A, P B, ΔT are the corresponding optimization variables or their functions. This inequality constraint defines the collaborative relationship that the process parameter combinations of relevant nozzle units must satisfy. Thus, the complete optimization model is defined as: Under the premise of satisfying the upper and lower limit constraints of each optimization variable itself and the above linear inequality constraints, find the optimization variable vector that maximizes the value of the objective function Q.

[0038] Secondly, a multi-objective evolutionary algorithm is used to solve the constructed constrained multi-objective optimization model. A set of solutions is selected from the obtained Pareto optimal solution set, and the independent process parameter values ​​of each nozzle unit corresponding to these solutions are used as the multi-nozzle cooperative process parameter combination to satisfy the cooperative constraint conditions. The specific solution process includes four core steps. The first step is algorithm initialization, where a multi-objective evolutionary algorithm is selected, such as a non-dominated sorting genetic algorithm with an elitist strategy. At this point, the necessary parameters for algorithm operation need to be set. The population size is set to W, for example, W equals 100, based on balancing computational efficiency and search space coverage. The maximum number of generations is set to Gmax, for example, Gmax equals 200, to ensure the algorithm has sufficient iteration opportunities while avoiding infinite loops. The crossover probability Pc controls the mixing degree of solutions, for example, set to 0.8. The mutation probability Pm maintains population diversity, for example, set to 0.1. Each candidate solution is encoded with a real number, and its gene sequence is an ordered arrangement of the values ​​of the optimization variable vector. The initial population is generated by random uniform sampling within the domain of all optimization variables. The second step is iterative evolution and fitness evaluation. In each generation of evolution, the optimization variable vector is decoded for each individual in the population, and its fitness is calculated. Fitness is determined by the objective function value and the penalty term for the degree of constraint violation. For a linear inequality constraint, the difference between the value of the left-hand side and the constant b of the right-hand side is calculated. If the difference is less than or equal to 0, the constraint is satisfied, and the penalty is 0; if the difference is greater than 0, the constraint is violated, and the penalty is the square of the difference multiplied by a penalty coefficient. The penalty coefficient K is set to be significantly larger than the typical value of the objective function Q, thereby effectively driving the search toward the feasible region. Specifically, by analyzing historical data or preliminary experiments, the common range of Q values ​​is between 1 and 10, and K is set to 1000. The total fitness F of an individual is equal to the objective function value Q minus the sum of all constraint penalties. The algorithm performs selection operations based on the fitness F, retaining high-quality individuals, and then generates offspring populations through crossover and mutation operations. This process is iterated until the maximum number of generations Gmax is reached. The third step is the extraction of the Pareto optimal solution set. After the algorithm terminates, fully feasible individuals whose sum of all constraint penalties is 0 are selected from the final generation population. These individuals may compete with each other on the two original quality dimensions. Through pairwise comparisons, solutions that are not completely dominated by any other feasible individuals on both the areal density uniformity U and thickness proximity C are selected; these solutions constitute the Pareto optimal solution set. The fourth step is the selection of the final cooperative parameter combination. From the Pareto optimal solution set, a final solution is selected according to a preset decision rule. The decision rule can be a single objective priority, such as selecting the solution with the largest thickness proximity C value; or it can be a comprehensive trade-off, such as calculating the weighted total score Q of each solution and selecting the solution with the largest Q value. The specific values ​​of each optimization variable obtained after decoding the gene sequence of the finally selected solution constitute a complete multi-nozzle cooperative process parameter combination that satisfies the cooperative constraints.Furthermore, if the Pareto optimal solution set is empty, meaning there is no completely feasible solution that simultaneously satisfies all constraints, then the individual with the smallest total constraint penalty value is selected as the alternative solution. This combination mathematically guarantees the optimization trend of global printing quality, while physically following the synergistic laws extracted from historical coupling patterns.

[0039] Example 2: Figure 2 A schematic diagram of a multi-objective collaborative optimization system for membrane material spraying and printing process parameters according to the present invention is provided. The multi-objective collaborative optimization system for membrane material spraying and printing process parameters includes: The data acquisition module is used to acquire the independent set of process parameters, the corresponding print quality dataset, and the print timing data for each printhead unit. The relationship judgment module is used to analyze the contribution ratio of the working time sequence of each printhead unit to the printing quality of the boundary area based on the print quality dataset and print timing data, and to determine the performance correlation between printhead units. The intensity calculation module is used to extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and to calculate the causal information transmission intensity between any two coupling influence modes. The community partitioning module is used to construct a directed weighted network of coupled influence patterns based on the strength of causal information transmission, and obtains multiple families of coupled patterns through community partitioning; The condition generation module is used to evaluate the matching degree between the current task condition characteristics and each coupled mode family, and to integrate the features of multiple related coupled mode families to generate a composite coupled influence mode as a collaborative constraint condition. The combined solution module is used to construct a multi-nozzle collaborative optimization model with independent process parameter sets as optimization variables, global printing quality index as optimization objective, and collaborative constraints, and solve for the multi-nozzle collaborative process parameter combination that satisfies the collaborative constraints.

[0040] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0041] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0042] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0043] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0044] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0045] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0046] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0047] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0049] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-objective collaborative optimization method for film coating and printing process parameters, characterized in that, include: S1. Obtain the independent process parameter set, corresponding print quality dataset, and print timing data for each printhead unit; S2. Based on the print quality dataset and print timing data, analyze the contribution ratio of the working timing of each printhead unit to the print quality of the boundary area, and determine the performance correlation between printhead units. S3. Extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and calculate the causal information transmission strength between any two coupling influence modes. S4. Construct a directed weighted network of coupling influence patterns based on the strength of causal information transmission, and obtain multiple families of coupling patterns through community partitioning; S5. Evaluate the matching degree between the current task condition characteristics and each coupling mode family, and integrate the characteristics of multiple related coupling mode families to generate a composite coupling influence mode as a cooperative constraint condition. S6. Using the set of independent process parameters as optimization variables, the global printing quality index as the optimization objective, and constructing a multi-nozzle collaborative optimization model with collaborative constraints, the combination of multi-nozzle collaborative process parameters that satisfy the collaborative constraints is obtained.

2. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 1, characterized in that, S1 includes: Obtain the independent set of process parameters for each printhead unit in historical printing tasks; Obtain the print quality dataset corresponding to each historical print task. The print quality dataset contains print quality measurements of the boundary areas. Obtain the printing timing data of each printhead unit when executing each historical printing task. The printing timing data includes the jet start and stop time sequence of each printhead unit.

3. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 1, characterized in that, S2 include: Based on the print quality dataset and print timing data, the print timing data of each printhead unit is time-aligned with the print quality measurement values ​​of the boundary region in the print quality dataset; A linear model is constructed using the printing timing data of each printhead unit after timing alignment and the printing quality measurement values ​​of the boundary area; the proportion of variance explained by the main effect of the working timing of each printhead unit to the total explained variance of the model is calculated by variance decomposition, which is taken as the contribution ratio of each printhead unit. Based on the contribution ratio of each nozzle unit, determine whether there is a performance correlation between any two nozzle units and the strength of the correlation.

4. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 1, characterized in that, S3 include: Based on performance correlation, the process parameter combinations corresponding to adjacent printhead units that have a coupled impact on the printing quality of the boundary area are identified; The identified process parameter combinations and the boundary area printing quality measurement values ​​are parametrically modeled to obtain the parametric relationship with process parameters as independent variables and boundary area printing quality measurement values ​​as dependent variables as the coupling influence mode. For any two coupled influence modes, the causal information transmission strength from one coupled influence mode to another is calculated based on the transfer entropy algorithm.

5. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 4, characterized in that, The calculation of causal information transmission strength based on the transfer entropy algorithm includes: treating the data sequences corresponding to two coupled influence modes as two random processes; calculating the conditional mutual information from the past state of the source random process to the current state of the target random process to obtain the transfer entropy value, which is used as the causal information transmission strength from the source coupled influence mode to the target coupled influence mode.

6. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 1, characterized in that, S4 includes: Each coupled influence pattern is treated as a network node, and the causal information transmission strength between any two coupled influence patterns is used as the weight of the directed edge between the corresponding nodes to construct a directed weighted network. The Louvain community detection algorithm is used to perform community partitioning on the constructed directed weighted network. The set of coupling influence patterns contained in each divided community is defined as a family of coupling patterns.

7. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 6, characterized in that, Community partitioning using the Louvain community discovery algorithm involves iteratively performing modularity optimization and community merging, continuously assigning nodes to communities that maximize network modularity, and merging communities into new nodes until modularity stabilizes. Finally, the set of coupling influence patterns within each stable community is defined as a family of coupling patterns.

8. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 1, characterized in that, S5 include: Based on the working characteristics of the current task and the characteristics of the coupling influence modes in each coupling mode family, calculate the matching degree between the current task and each coupling mode family. Based on the matching degree, select the families of coupling patterns with the highest matching degree as the relevant coupling pattern families; The features of each selected family of related coupling modes are weighted and fused according to their corresponding matching degree to generate a composite coupling influence mode. The generated complex coupling influence pattern is used as a cooperative constraint condition.

9. The multi-objective collaborative optimization method for film coating and printing process parameters according to claim 1, characterized in that, S6 include: Using the independent process parameters of each nozzle unit in the independent process parameter set as optimization variables, maximizing the global printing quality index as the optimization objective, and the parameter relationship characterized by the complex coupling influence mode as the inequality constraint, a constrained multi-objective optimization model is constructed. A multi-objective evolutionary algorithm is used to solve the constructed constrained multi-objective optimization model. A set of solutions is selected from the obtained Pareto optimal solution set, and the independent process parameter values ​​of each nozzle unit corresponding to these solutions are used as the multi-nozzle collaborative process parameter combination that satisfies the collaborative constraint conditions.

10. A multi-objective collaborative optimization system for film coating and printing process parameters, used to implement the multi-objective collaborative optimization method for film coating and printing process parameters as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire the independent set of process parameters, the corresponding print quality dataset, and the print timing data for each printhead unit. The relationship judgment module is used to analyze the contribution ratio of the working time sequence of each printhead unit to the printing quality of the boundary area based on the print quality dataset and print timing data, and to determine the performance correlation between printhead units. The intensity calculation module is used to extract the coupling influence mode of the process parameter combination of adjacent printhead units on the printing quality of the boundary area from the performance correlation, and to calculate the causal information transmission intensity between any two coupling influence modes. The community partitioning module is used to construct a directed weighted network of coupled influence patterns based on the strength of causal information transmission, and obtains multiple families of coupled patterns through community partitioning; The condition generation module is used to evaluate the matching degree between the current task condition characteristics and each coupled mode family, and to integrate the features of multiple related coupled mode families to generate a composite coupled influence mode as a collaborative constraint condition. The combined solution module is used to construct a multi-nozzle collaborative optimization model with independent process parameter sets as optimization variables, global printing quality index as optimization objective, and collaborative constraints, and solve for the multi-nozzle collaborative process parameter combination that satisfies the collaborative constraints.

Images