A method and device for preparing multi-layer composite thin film material with multi-node coordination

Abstract

The present invention discloses a method and device for preparing a multi-layer composite thin film material with multi-node collaboration, and relates to the technical field of material preparation. The method comprises: constructing a process graph structure by taking the process flow of preparing the multi-layer composite thin film as a graph node, taking the physical parameter connection between the process flows as an edge, setting constraints according to the timing relationship and fitting them into the edge, and constraining the delay time from the node to the downstream node. The process sequence is determined according to the timing constraint, and each node transmits and aggregates the upstream feature information through the edge, generates a new node feature in combination with its own features, and then obtains the node preparation parameters according to the new feature analysis. The technical problem that the process flow nodes in the existing multi-layer composite thin film preparation are independent and lack of timing constraints, which reduces the preparation uniformity and product quality stability is solved. Through the high degree of collaboration and precise control of the multi-layer composite thin film preparation process, the technical effect of improving the preparation uniformity and product quality stability is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of material preparation, and in particular to a method and device for preparing a multi-node coordinated multi-layer composite thin film material. Background Art

[0002] In the scenario of multi-layer composite thin film material preparation, the coordinated control and precise grasp of each process link are particularly prominent. Achieving efficient process coordination and precise parameter control has become a crucial link in improving the quality of multi-layer composite thin film materials. Traditional multi-layer composite thin film material preparation methods are often relatively independent and lack systematicity. Each process node is carried out independently. There is insufficient grasp of the physical parameter connection between different process links, and there is a lack of consideration of the timing relationship between process flows. There are unreasonable situations in resource allocation (i.e. process parameter adjustment and process coordinated control), resulting in unstable product quality and low production efficiency. The formulation of the preparation plan is relatively simple and fixed, and cannot cope well with the complex and changeable actual preparation situation.

[0003] In the current related technologies, each process node in the preparation of multilayer composite films is independent and lacks timing constraints, which reduces the technical problems of preparation uniformity and product quality stability. Summary of the invention

[0004] The present application provides a multi-node collaborative multi-layer composite thin film material preparation method and device, uses the multi-layer composite thin film preparation process flow as a graph node to construct a process graph structure, uses the physical parameter connection between the process flows as an edge, sets constraints according to the timing relationship and fits them into the edge, and constrains the delay time from the node to the downstream node. The process sequence is determined according to the timing constraint, each node transmits upstream feature information through the edge and aggregates, generates new node features based on its own features, and then obtains node preparation parameters based on the new feature analysis. Through the high degree of coordination and precise control of the multi-layer composite thin film preparation process, the technical effect of improving preparation uniformity and product quality stability is achieved.

[0005] The present application provides a method for preparing a multi-node coordinated multi-layer composite thin film material, comprising: The preparation process flow of the multilayer composite film is taken as a graph node, each process flow node is taken as a vertex, and the association relationship between the nodes is taken as an edge to construct a process graph structure, wherein the association relationship is the physical parameter connection between the process flows; timing constraints are set according to the timing relationship of the preparation process flow, and the timing constraints are fitted into the edges, and the timing constraints are used to constrain the delay time from each node to the downstream node; the process sequence is determined according to the timing constraints, and each node transmits the feature information of the adjacent upstream nodes through the edges in the process graph structure, and aggregates the received feature information, and each node uses an update function to combine the aggregated feature information with its own features to generate new node features; the preparation control parameters are parsed according to the new node features to obtain the node preparation parameters.

[0006] The present application also provides a multi-node coordinated multi-layer composite thin film material preparation device, comprising: A process graph structure construction module, the process graph structure construction module is used to take the preparation process flow of the multilayer composite film as a graph node, each process flow node as a vertex, and the association relationship between the nodes as an edge to construct a process graph structure, wherein the association relationship is the physical parameter connection between the process flows; a constraint fitting module, the constraint fitting module is used to set timing constraints according to the timing relationship of the preparation process flow, fit the timing constraints to the edges, and the timing constraints are used to constrain the delay time from each node to the downstream node; a node feature generation module, the node feature generation module is used to determine the process sequence according to the timing constraints, through the edges in the process graph structure, each node transmits the feature information of the adjacent upstream nodes, and aggregates the received feature information, each node uses an update function to combine the aggregated feature information with its own features to generate new node features; a node preparation parameter acquisition module, the node preparation parameter acquisition module is used to perform preparation control parameter analysis according to the new node features to obtain node preparation parameters.

[0007] The multi-node collaborative multi-layer composite thin film material preparation method and device proposed in this application first constructs a process graph structure by taking the multi-layer composite thin film preparation process flow as a graph node, and takes the physical parameter connection between the process flows as an edge. Constraints are set according to the timing relationship and fitted into the edge to constrain the delay time from the node to the downstream node. The process sequence is determined according to the timing constraints. Each node transmits upstream feature information through the edge and aggregates it. New node features are generated based on its own features, and then node preparation parameters are obtained based on the new feature analysis. Through the high degree of coordination and precise control of the multi-layer composite thin film preparation process, the technical effect of improving preparation uniformity and product quality stability is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the device according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0009] Figure 1 A schematic diagram of a process for preparing a multi-node coordinated multi-layer composite thin film material provided in an embodiment of the present application; Figure 2 A schematic structural diagram of a multi-node coordinated multi-layer composite thin film material preparation device provided in an embodiment of the present application.

[0010] Explanation of the reference numerals: process diagram structure building module 10 , constraint fitting module 20 , node feature generating module 30 , node preparation parameter acquiring module 40 . DETAILED DESCRIPTION

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0012] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0013] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0014] The present application embodiment provides a method for preparing a multi-node coordinated multi-layer composite thin film material, such as Figure 1 As shown, the method includes: Step S100, the preparation process flow of the multilayer composite film is taken as a graph node, each process flow node is taken as a vertex, and the association relationship between the nodes is taken as an edge to construct a process graph structure, wherein the association relationship is the physical parameter relationship between the process flows. Specifically, when constructing the process graph structure of the preparation of the multilayer composite film material, the entire preparation process flow is taken as a graph node, wherein each process flow node such as unwinding the PET base film, slit coating, coating pre-curing oven, coating structure forming, UV curing coating, protective film compounding, cutting waste edges and finished product winding is taken as a vertex in the graph, and the association relationship between the nodes, that is, the physical parameter relationship between the process flows is taken as an edge. For example, from unwinding the PET base film to slit coating, the parameters such as the tension, speed and thickness of the base film will affect the slit coating effect, and the edge between the two vertices represents this physical parameter relationship. In this way, a process graph structure that clearly displays the physical parameter relationship of each process flow node is constructed, providing an intuitive model for subsequent analysis and optimization.

[0015] In one possible implementation, the preparation process flow of the multilayer composite film is taken as a graph node, each process flow node is taken as a vertex, and the association relationship between the nodes is taken as an edge to construct a process graph structure, wherein the association relationship is the physical parameter connection between the process flows, and step S100 further includes step S110, determining the dependency relationship and the dependency strength between the process flow nodes according to the physical principles of the process flow nodes and historical sample data. Specifically, the physical principles of the process nodes are analyzed. For the preparation process of multi-layer composite film materials, each process node has its specific physical effects and influencing factors. For example, in the node of unwinding PET base film, the physical parameters such as the tension and speed of the base film will affect the subsequent slit coating node; in the node of coating pre-curing oven, parameters such as temperature and time will affect the curing degree of the coating, and then affect the subsequent coating structure forming node. The historical sample data are analyzed. The historical sample data contains various parameter records in the past multi-layer composite film preparation process, such as the operating parameters of different process nodes, product quality data, etc. Through statistical analysis and machine learning algorithm processing of the historical sample data, the correlation between different process nodes can be found. The physical principles and historical sample data are combined to determine the dependency between the process nodes. For example, it is found that the coating thickness of the slit coating node is largely dependent on the tension and speed of the unwinding PET base film node, and the temperature of the coating pre-curing oven node. At the same time, the dependency strength is determined. For example, through quantitative analysis, it is determined that the dependency strength of the unwinding PET base film node on the slit coating node is high, while the dependency strength of the coating pre-curing oven node on the slit coating node is medium.

[0016] Step S120, fitting the change attenuation according to the dependency strength, and determining the segmentation threshold. Specifically, fitting the change attenuation according to the dependency strength determined previously, taking into account that as the process progresses, the influence of the upstream node on the downstream node may gradually decay, for example, the influence of the upstream node at a distance may be relatively small on the current node, by mathematically fitting the change in dependency strength, the attenuation law can be found out, and the segmentation threshold is determined. The segmentation threshold is used to distinguish which upstream nodes have a greater influence on the current node and which have a smaller influence. For example, a dependency strength value is set as the segmentation threshold. If the dependency strength of a certain upstream node on the current node is higher than this threshold, it is considered that the node has an important influence on the current node; if it is lower than this threshold, it is considered that the influence is small and can be ignored.

[0017] Step S130, using the segmentation threshold to determine the dependent process nodes, and determine the number of split node layers. Specifically, the segmentation threshold is used to determine the dependent process nodes, and the dependency strength of each process node with other nodes is compared with the segmentation threshold. If the dependency strength is higher than the segmentation threshold, the node is determined as a dependent process node. For example, in the slit coating node, after comparison, it is found that the unwinding PET base film node and the coating pre-curing oven node are dependent process nodes, because their dependency strength on the slit coating node is higher than the segmentation threshold. The number of split node layers is determined, and the distance from the current node to the farthest dependent process node is calculated based on the determined dependent process nodes. This distance is the number of split node layers. For example, if the farthest dependent process node of the slit coating node is the unwinding PET base film node, and a coating pre-curing oven node is passed in the middle, then the number of split node layers is two.

[0018] Step S140, setting the number of network layers according to the number of split node layers, and setting the number of network layers for edge transmission for the process graph structure. Specifically, the number of network layers is set according to the number of split node layers. If the number of split node layers is small, it means that the range of upstream node features to be transmitted is small, and a small number of network layers can be set. For example, if it is desired to transmit only the nearest 1-2 upstream node features, and the number of split node layers is one or two layers, a small number of network layers (such as 1-2 layers) can be set. If the number of split node layers is large and it is necessary to transmit farther upstream node features, the number of network layers is increased, and the number of network layers for edge transmission is set for the process graph structure. In the constructed process graph structure, the edge transmission rules are set according to the determined number of network layers. For example, if two network layers are set, then at the slit coating node, feature information from the unwinding PET base film node and the coating pre-curing oven node will be received, and transmitted and aggregated through the edges, ensuring that in the subsequent determination of the preparation control parameters, the influence of the upstream nodes can be accurately considered, and the coordination and accuracy of the preparation process can be improved.

[0019] In a possible implementation, the variation attenuation is fitted according to the dependency strength to determine the segmentation threshold, and step S120 further includes step S121, determining the discreteness of the variation according to the dependency strength. Specifically, the dependency strength between the process nodes has been determined, and the dependency strength reflects the degree of mutual influence between different process nodes. The discreteness of the variation of the dependency strength is calculated by using the variance, standard deviation, etc., and the formula ,in represents the number of samples, represents the change in each dependency strength, It represents the average value of the change in dependency intensity. The formula can be used to get the value of the discreteness. If the discreteness is large, it means that the dependency relationship between different nodes changes dramatically. The impact of some nodes on the downstream process may fluctuate greatly, while the impact of other nodes is relatively weak.

[0020] Step S122, determine the center point according to the dispersion. Specifically, after obtaining the dispersion, determine the center point according to the dispersion. The center point can be determined in a variety of ways, for example, the average value of the dependence strength change can be taken as the center point. If the dispersion is large, it means that the dependence strength change is more dispersed. At this time, the determination of the center point may need to consider more factors, such as more robust statistics such as the median. If the dispersion is small, the average value may be a more appropriate center point.

[0021] Step S123, obtain the adjustment constant according to the half amount of the maximum dependency strength and the second largest dependency strength change. Specifically, determine the maximum dependency strength and the second largest dependency strength, find the maximum value and the second largest value among all dependency strengths, calculate the half amount of the maximum dependency strength and the second largest dependency strength change, first calculate the difference between the maximum dependency strength and the second largest dependency strength, and then take half of the difference as the adjustment constant, which reflects a relative amplitude of the dependency strength change and is used to adjust the segmentation threshold.

[0022] Step S124, according to the dependency strength corresponding to the center point and the adjustment constant, the segmentation threshold is obtained. Specifically, the segmentation threshold is obtained according to the dependency strength corresponding to the center point and the adjustment constant determined previously. The calculation formula of the segmentation threshold can be segmentation threshold = dependency strength corresponding to the center point + adjustment constant. The segmentation threshold will be used to determine which upstream nodes have a greater impact on the current node, thereby playing a key role in the subsequent process graph structure edge transfer network layer number setting.

[0023] In a possible implementation, the segmentation threshold is obtained according to the dependency strength corresponding to the center point and the adjustment constant, and step S124 further includes step S1241, where the expression of the segmentation threshold is: ,in, is the dependence strength corresponding to the center point, is the adjustment constant, ,in, is the maximum dependency strength, Specifically, in the process of determining the segmentation threshold, the dependency intensity corresponding to the center point is first determined by analyzing the physical principles of the process nodes and historical sample data. . Find the maximum dependency strength among all the dependency strengths between process nodes and the second largest dependency strength , according to the formula Calculate the adjustment constant , the expression of segmentation threshold is The split threshold is used to distinguish the degree of influence of different process flow nodes on specific nodes. Upstream nodes above the split threshold are considered to have a greater impact on the current node, while those below the split threshold have a smaller impact. This provides an important basis for subsequent operations such as determining dependent process nodes, splitting node layers, and setting the number of network layers.

[0024] Step S200, set the timing constraint conditions according to the timing relationship of the preparation process flow, fit the timing constraint conditions to the edge, and the timing constraint conditions are used to constrain the delay time from each node to the downstream node. Specifically, in the preparation process of the multilayer composite film, first clarify the timing relationship of the process flow, each process node is carried out in sequence, set the timing constraint conditions according to the actual situation and requirements, determine the delay time range from each node to the downstream node, the delay time can be determined based on historical data, process standards and actual production experience, and then fit the timing constraint conditions to the edge. In the constructed process graph structure, the edge represents the connection and dependency relationship between nodes. After adding the delay time constraint information, when setting the number of edge transmission network layers and feature transmission, the information transmission between nodes can be adjusted and optimized according to the constraint conditions. When determining the preparation control parameters, the influence of the upstream node on the current node is considered and combined with the timing constraint conditions. If the delay time of the upstream node is too long, it will affect the adjustment of the control parameters of the current node. Fitting the timing constraint conditions to the edge can effectively constrain the delay time from the node to the downstream node, thereby improving the stability and controllability of the preparation process.

[0025] Step S300, determine the process sequence according to the timing constraints, and through the edges in the process graph structure, each node transmits the feature information of the adjacent upstream nodes, and aggregates the received feature information. Each node uses an update function to combine the aggregated feature information with its own features to generate new node features. Specifically, in the preparation process of the multilayer composite film, first determine the process sequence according to the timing constraints to ensure that the preparation process is carried out in an orderly manner. Using the edges in the process graph structure, each node transmits the feature information of the adjacent upstream nodes, and then aggregates the received feature information. Through the aggregation operation, the information from different upstream nodes is integrated into a unified feature representation. Each node uses an update function to combine the aggregated feature information with its own features to generate new node features. For example, when the unwinding tension is large, the coating node can aggregate the information and use the update function to adjust the coating thickness in time, improve the preparation uniformity, realize information interaction and collaborative optimization between each process node, and improve the preparation quality and efficiency.

[0026] In a possible implementation, the process sequence is determined according to the timing constraints, and each node transmits the feature information of the adjacent upstream nodes through the edges in the process graph structure, and aggregates the received feature information. Each node uses an update function to combine the aggregated feature information with its own features to generate new node features. Step S300 further includes step S310, setting the influence weight of the edge transfer network layer. Specifically, in the process graph structure for the preparation of multi-layer composite films, the edges represent the connection and dependency between nodes. In order to better control the transmission and aggregation of features between nodes, it is first necessary to set the influence weight of the edge transfer network layer. The influence weight reflects the importance of upstream nodes at different levels to downstream nodes. For example, in some processes, the adjacent upstream nodes may have a greater impact on the current node, while the upstream nodes at farther levels have a relatively smaller impact. By reasonably setting the influence weight, the interaction between nodes in the actual process flow can be more accurately simulated.

[0027] Step S320, according to the impact weight, the characteristic information is transmitted downward in sequence from the most upstream node according to the hierarchical relationship of the process flow and accumulated, and is transmitted to the downstream node through multiple aggregations to obtain aggregated characteristic information. Specifically, after determining the impact weight, the characteristic information is transmitted from the most upstream node according to the hierarchical relationship of the process flow. Since the process flow has a clear sequence, starting from the most upstream node can ensure that the source of the information is accurate. During the transmission process, according to the impact weight, the characteristic information of the upstream node is transmitted downward in sequence and accumulated. Each time it is transmitted to a node, the node will aggregate the characteristic information received from different upstream nodes. Through multiple aggregation operations, the characteristic information is continuously integrated and optimized, and finally transmitted to the downstream node to obtain aggregated characteristic information. The aggregated characteristic information contains the status and characteristics of multiple upstream nodes, which can provide a comprehensive reference for the feature update of the downstream node.

[0028] Step S330, by updating the function: , combining the aggregate feature information with its own features to obtain the new node features; wherein, is the new feature vector of node j at time t+1, is the activation function, is the influence weight of the upstream node on the downstream node, The set of upstream nodes at the edge transfer network level, is the feature of upstream node i at time t, is the feature vector of the current node j at time t, is a trainable weight matrix used to perform linear transformation on the aggregated features. Specifically, is the new feature vector of node j at time t+1. The feature vector represents the final state of node j after feature transfer and update. It contains the current state information of the node in the process flow, such as tension, speed, temperature, thickness and other process parameters. Through feature transfer and update, the features of the node are constantly adjusted and optimized. The activation function is a nonlinear function used to perform nonlinear transformation on node features. Common activation functions include ReLU, Sigmoid, and Tanh. The function of the activation function is to perform nonlinear mapping on the features after linear transformation, so that the model can capture more complex patterns and relationships, and help nodes to perform nonlinear adjustments during feature transmission and updating. For example, the ReLU function sets negative values ​​to zero and keeps positive values ​​unchanged; the Sigmoid function maps the input value to the [0,1] interval; the Tanh function maps the input value to the [-1,1] interval. is the influence weight of the upstream node on the downstream node, which determines the contribution of different upstream nodes to the current node. is the set of upstream nodes at the edge transfer network level, including all upstream nodes that have an impact on node j. is the feature of upstream node i at time t, representing the state information of the upstream node. The current node j’s own feature vector at time t reflects the initial state of node j at the current time. It is a trainable weight matrix used to perform linear transformation on the aggregated features. By training this weight matrix, it can better adapt to different process flows and data distributions. First, calculate , that is, the features of all upstream nodes are multiplied by the corresponding influence weights and then summed, which means that the feature information of the upstream nodes is aggregated, and the contribution of different upstream nodes to the current node is considered. The aggregated upstream feature information is combined with the current node's own feature vector Add together and get . Input the result of linear transformation into the activation function , get the new feature vector of node j at time t+1 , the aggregated feature information is combined with its own features to achieve the update and optimization of node features.

[0029] Step S400, the preparation control parameters are parsed according to the new node characteristics to obtain the node preparation parameters. Specifically, after obtaining the new node characteristics, they are analyzed and interpreted to extract key information related to preparation control, such as inferring the preparation control parameters such as heating power or coating speed that need to be adjusted based on the temperature and thickness characteristics of the node. This process may involve complex mathematical models and algorithms or rule derivation based on historical data and experience. The preparation control parameters are parsed by comprehensively considering multiple node characteristics and analyzing the relationship between different process links, and finally the node preparation parameters are obtained to guide the actual preparation process, ensure that the quality and performance of the multi-layer composite film meet the requirements, and it may be necessary to continuously adjust and optimize the preparation control parameters according to the new node characteristics to adapt to different production conditions and demand changes.

[0030] In one possible implementation, preparation control parameters are parsed according to new node features to obtain node preparation parameters, and step S400 further includes step S410, constructing a mapping function, wherein the mapping function is obtained by learning the conversion relationship using a historical training data set through a machine learning model, and the historical training data set includes a feature vector, preparation control parameters and their mapping relationship, wherein the mapping function is used to convert the feature vector into the preparation control parameter. Specifically, historical training data sets are collected, which contain feature vectors, preparation control parameters and the mapping relationship between them. The feature vectors reflect the state information of each node in the process flow, such as process parameters such as tension, speed, temperature, thickness, etc. The preparation control parameters are used to adjust and control the specific parameters of the preparation process, such as heating power, coating speed, pressure, etc. Through the analysis of historical data, the corresponding relationship between different feature vectors and preparation control parameters can be understood. The historical training data sets are used to learn the conversion relationship through machine learning models, and appropriate machine learning algorithms such as linear regression, neural networks, decision trees, etc. can be selected to learn the complex relationship between feature vectors and preparation control parameters from historical data. For example, for cases where the linear relationship is more obvious, a linear regression algorithm can be used; for cases where the relationship is more complex, nonlinear models such as neural networks can be used to finally construct a mapping function. The mapping function can convert feature vectors into preparation control parameters, providing a basis for subsequent control parameter analysis.

[0031] Step S420, taking the new node feature as an input variable, performing feature mapping through the mapping function, and obtaining the preparation control parameter as the node preparation parameter. Specifically, through the previous feature transfer and update process, the latest feature of the current node, that is, the new node feature, is obtained. The new node feature contains the current state information of the node in the process flow after a series of processing and updates. The new node feature is taken as an input variable, and feature mapping is performed through the constructed mapping function. According to the relationship between the feature and the control parameter, a linear mapping method or a nonlinear mapping method can be selected. If there is a relatively linear relationship between the feature and the control parameter, a linear mapping method can be used, that is, the formula ,in is the control parameter vector of node j, is the linear mapping matrix from features to control parameters, is the bias term. For example, the characteristics of the coating node (such as thickness and speed) can be directly resolved into the required control parameters such as coating thickness and coating speed through linear mapping; the tension control parameter can be linearly adjusted according to the aggregated characteristic tension information. If the relationship between the characteristics and the control parameters is more complex, a nonlinear mapping function can be used, such as a neural network structure such as a multilayer perceptron, that is, the formula ,in It is a nonlinear mapping function. For example, in some complex processes, such as the effect of temperature on coating thickness, nonlinear mapping may be required. When the coating is too thick, the temperature of the curing process may need to be adjusted nonlinearly. At this time, using a machine learning model to fit historical data can predict the optimal control parameters under different feature combinations, and adjust the preparation process in real time according to the control parameters parsed by the mapping function. For example, the operating speed of the coating equipment is adjusted according to the obtained coating speed control parameters, and the power of the heating device is adjusted according to the heating power control parameters. By adjusting the preparation process in real time, product quality can be improved, the production process can be optimized, new production data can be continuously collected, historical training data sets can be updated, and mapping functions can be further optimized to achieve continuous improvement. By constructing a mapping function and mapping the new node features through the mapping function, the preparation control parameters can be obtained as node preparation parameters to achieve precise control of the preparation process of multilayer composite films.

[0032] In one possible implementation, the new node feature is used as an input variable, feature mapping is performed through the mapping function, and the preparation control parameter is obtained as the node preparation parameter. Step S420 further includes step S421, setting the equipment control parameter limit of each process flow node, and using the equipment control parameter limit to determine whether the node preparation parameter meets the limit requirement. When it does not meet the limit requirement, a hard constraint is implemented to assign the limit to the node preparation parameter constraint. Specifically, according to the requirements of the process flow and the performance characteristics of the equipment, the equipment control parameter limits of each process flow node are set. The limits include upper and lower limits, such as the maximum allowable value of the coating thickness, the range of the heating temperature, etc. The determination of the equipment control parameter limits needs to take into account factors such as product quality requirements, safe operation of the equipment, and process stability. The set equipment control parameter limits are used to determine whether the node preparation parameters obtained through the previous steps meet the limit requirements. The node preparation parameters are compared with the equipment control parameter limits. If the node preparation parameters are within the limit range, it means that the parameters meet the requirements; if the node preparation parameters exceed the limit range, the requirements are not met. When the node preparation parameters do not meet the limit requirements, hard constraints are implemented. For parameters that exceed the upper and lower limits, they can be directly adjusted to the boundary values. For example, if the mapped coating thickness exceeds the maximum allowable thickness Pmax, it is corrected to Pmax. This ensures that the node preparation parameters are within the range that the equipment can bear and ensures the safe and stable operation of the process.

[0033] Step S422, determine the deviation control parameter according to the node preparation parameter and the equipment control parameter limit, perform compensation analysis on the deviation control parameter according to the compensation parameter, and add the compensation parameter and the compensation control amount. Specifically, determine the deviation control parameter according to the comparison result between the node preparation parameter and the equipment control parameter limit. The deviation control parameter refers to the difference between the node preparation parameter and the equipment control parameter limit. For example, if the node preparation parameter is a coating thickness of 0.007mm, and the equipment control parameter limit is 0.006mm, then the deviation control parameter is 0.001mm. Perform compensation analysis on the deviation control parameter according to the compensation parameter. The compensation parameter is determined based on process experience and historical data and is used to compensate for the deviation between the node preparation parameter and the equipment control parameter limit. For example, in the preparation process of a multilayer composite film, due to process limitations, the coating thickness is adjusted from the ideal 0.007mm to 0.006mm, which affects the coating drying effect. To compensate for this deviation, you can make Using methods such as temperature compensation and time compensation, according to the compensation model, the drying temperature is increased by 5°C to compensate for the reduction in coating thickness. By increasing the drying temperature, the drying speed of the coating can be accelerated to ensure that the material can be fully dried even with a thinner coating. Time compensation: extend the drying time by 10 seconds to ensure that the material can be fully dried even with a thinner coating. By extending the drying time, the residence time of the material in the drying equipment can be increased to improve the drying effect. After determining the compensation parameters and compensation control amounts, add them to the deviation control parameters. The compensation control amount refers to the specific adjustment value determined based on the compensation parameters. For example, if the temperature compensation is to increase the drying temperature by 5°C, the compensation control amount is 5°C. Adding the compensation control amount to the deviation control parameters can obtain the adjusted deviation control parameters.

[0034] Step S423, the compensation parameters and compensation control amounts are integrated into the node preparation parameters, and the node preparation parameters are updated. Specifically, the compensation parameters and compensation control amounts are integrated into the node preparation parameters, and the node preparation parameters are updated. The adjusted deviation control parameters are combined with the node preparation parameters to obtain updated node preparation parameters. The updated node preparation parameters not only take into account the requirements of the equipment control parameter limits, but also are adjusted through compensation parameters and compensation control amounts, so as to better adapt to the process requirements and equipment performance. By setting the equipment control parameter limits, determining the deviation control parameters, performing compensation analysis, and updating the node preparation parameters, the multilayer composite film preparation process can be accurately controlled, and the product quality and process stability can be improved.

[0035] The embodiment of the present application uses the multi-layer composite film preparation process as a graph node to construct a process graph structure, uses the physical parameter connection between the process flows as an edge, sets constraints according to the timing relationship and fits them into the edge, and constrains the delay time from the node to the downstream node. The process sequence is determined according to the timing constraints, and each node transmits upstream feature information through the edge and aggregates it, generates new node features based on its own features, and then obtains node preparation parameters based on the new feature analysis. Through the high degree of coordination and precise control of the multi-layer composite film preparation process, the technical effect of improving preparation uniformity and product quality stability is achieved.

[0036] In the above, refer to Figure 1 A method for preparing a multi-node coordinated multi-layer composite thin film material according to an embodiment of the present invention is described in detail. Figure 2 A multi-node coordinated multi-layer composite thin film material preparation device according to an embodiment of the present invention is described.

[0037] A multi-node collaborative multi-layer composite thin film material preparation device according to an embodiment of the present invention is used to solve the technical problem that each process node in the existing multi-layer composite thin film preparation is independent and lacks timing constraints, which reduces the preparation uniformity and product quality stability. Through the high degree of coordination and precise control of the multi-layer composite thin film preparation process, the technical effect of improving the preparation uniformity and product quality stability is achieved. A multi-node collaborative multi-layer composite thin film material preparation device includes: a process diagram structure construction module 10, a constraint fitting module 20, a node feature generation module 30, and a node preparation parameter acquisition module 40.

[0038] The process graph structure building module 10 is used to construct a process graph structure by taking the preparation process flow of the multilayer composite film as a graph node, each process flow node as a vertex, and the association relationship between the nodes as an edge, wherein the association relationship is the physical parameter connection between the process flows.

[0039] The constraint fitting module 20 is used to set timing constraints according to the timing relationship of the manufacturing process flow, and fit the timing constraints to the edges. The timing constraints are used to constrain the delay time from each node to the downstream node.

[0040] The node feature generation module 30 is used to determine the process sequence according to the timing constraints. Through the edges in the process graph structure, each node transmits the feature information of the adjacent upstream nodes and aggregates the received feature information. Each node uses an update function to combine the aggregated feature information with its own features to generate new node features.

[0041] The node preparation parameter acquisition module 40 is used to perform preparation control parameter analysis according to new node characteristics to obtain node preparation parameters.

[0042] The specific configuration of the process graph structure construction module 10 will be described in detail below. As described above, the preparation process flow of the multilayer composite film is taken as a graph node, each process flow node is taken as a vertex, and the association relationship between the nodes is taken as an edge to construct a process graph structure, wherein the association relationship is a physical parameter connection between the process flows, and the process graph structure construction module 10 further includes: a dependency determination unit, the dependency determination unit is used to determine the dependency relationship and its dependency strength between the process flow nodes according to the physical principle of the process flow nodes and historical sample data; a segmentation threshold determination unit, the segmentation threshold determination unit is used to perform a segmentation threshold fitting according to the dependency strength to determine the segmentation threshold; a segmentation node layer determination unit, the segmentation node layer determination unit is used to determine the dependent process nodes using the segmentation threshold and determine the number of segmentation node layers; a network layer setting unit, the network layer setting unit is used to set the number of network layers according to the number of segmentation node layers, and perform edge transfer network layer setting on the process graph structure.

[0043] Among them, the segmentation threshold is fitted according to the change attenuation amount of the dependency strength to determine the segmentation threshold, and the segmentation threshold determination unit further includes: a discreteness determination subunit, the discreteness determination subunit is used to determine the discreteness of the change amount according to the dependency strength; a center point determination subunit, the center point determination subunit is used to determine the center point according to the discreteness; an adjustment constant acquisition subunit, the adjustment constant acquisition subunit is used to obtain the adjustment constant according to half of the maximum dependency strength and the second largest dependency strength change amount; a segmentation threshold acquisition subunit, the segmentation threshold acquisition subunit is used to obtain the segmentation threshold according to the dependency strength corresponding to the center point and the adjustment constant.

[0044] According to the dependency strength corresponding to the center point and the adjustment constant, the segmentation threshold is obtained, and the segmentation threshold acquisition subunit further includes: a segmentation threshold expression composition micro-unit, and the segmentation threshold expression composition micro-unit is used for the expression of the segmentation threshold: ,in, is the dependence strength corresponding to the center point, is the adjustment constant, ,in, is the maximum dependency strength, This is the highest level of dependence.

[0045] The specific configuration of the node feature generation module 30 will be described in detail below. As described above, the process sequence is determined according to the timing constraints, and each node transmits the feature information of the adjacent upstream nodes through the edges in the process graph structure, and aggregates the received feature information. Each node uses an update function to combine the aggregated feature information with its own features to generate new node features. The node feature generation module 30 further includes: an influence weight setting unit, which is used to set the influence weight of the edge transmission network layer; an aggregate feature information acquisition unit, which is used to transmit and accumulate feature information from the upstream node in sequence according to the hierarchical relationship of the process flow according to the influence weight, and transmit it to the downstream node through multiple aggregations to obtain aggregate feature information; a new node feature acquisition unit, which is used to pass through the update function: , combining the aggregate feature information with its own features to obtain the new node features; wherein, is the new feature vector of node j at time t+1, is the activation function, is the influence weight of the upstream node on the downstream node, The set of upstream nodes at the edge transfer network level, is the feature of upstream node i at time t, is the feature vector of the current node j at time t, It is a trainable weight matrix used to perform linear transformation on the aggregated features.

[0046] The specific configuration of the node preparation parameter acquisition module 40 will be described in detail below. As described above, the preparation control parameter is parsed according to the new node features to obtain the node preparation parameters. The node preparation parameter acquisition module 40 further includes: a mapping function construction unit, the mapping function construction unit is used to construct a mapping function, the mapping function is obtained by learning the conversion relationship using a historical training data set through a machine learning model, the historical training data set includes a feature vector, a preparation control parameter and its mapping relationship, wherein the mapping function is used to convert the feature vector into a preparation control parameter; a node preparation parameter acquisition unit, the node preparation parameter acquisition unit is used to use the new node feature as an input variable, perform feature mapping through the mapping function, and obtain the preparation control parameter as the node preparation parameter.

[0047] Among them, the new node feature is taken as the input variable, and the feature mapping is performed through the mapping function to obtain the preparation control parameter as the node preparation parameter. The node preparation parameter acquisition unit further includes: an equipment control parameter limit setting subunit, the equipment control parameter limit setting subunit is used to set the equipment control parameter limit of each process flow node, and use the equipment control parameter limit to determine whether the node preparation parameter meets the limit requirement. When it does not meet the limit requirement, a hard constraint is implemented to assign the limit to the node preparation parameter constraint; a deviation control parameter determination subunit, the deviation control parameter determination subunit is used to determine the deviation control parameter according to the node preparation parameter and the equipment control parameter limit, perform compensation analysis on the deviation control parameter according to the compensation parameter, and add compensation parameters and compensation control quantities; a preparation parameter update subunit, the preparation parameter update subunit is used to merge the compensation parameters and compensation control quantities into the node preparation parameters to update the node preparation parameters.

[0048] A multi-node coordinated multi-layer composite thin film material preparation device provided in an embodiment of the present invention can execute a multi-node coordinated multi-layer composite thin film material preparation method provided in any embodiment of the present invention, and has corresponding functional modules and beneficial effects of the execution method.

[0049] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0050] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for preparing a multi-node coordinated multi-layer composite thin film material, characterized in that: include: The preparation process of multilayer composite film is taken as a graph node, each process node is taken as a vertex, and the association relationship between nodes is taken as an edge to construct a process graph structure, where the association relationship is the physical parameter connection between the process flows. Setting timing constraints according to the timing relationship of the manufacturing process flow, fitting the timing constraints to the edges, and the timing constraints are used to constrain the delay time from each node to the downstream node; Determine the process sequence according to the timing constraints, and transmit the feature information of the adjacent upstream nodes through the edges in the process graph structure, and aggregate the received feature information. Each node combines the aggregated feature information with its own features using an update function to generate new node features; The preparation control parameters are parsed according to the new node characteristics to obtain the node preparation parameters.

2. The method for preparing a multi-node coordinated multi-layer composite thin film material according to claim 1, characterized in that: Also includes: Determine the dependency relationship and dependency strength between process nodes based on the physical principles of process nodes and historical sample data; Fitting the change attenuation amount according to the dependency strength to determine the segmentation threshold; Determine the dependent process nodes and the number of split node layers by using the split threshold; The number of network layers is set according to the number of split node layers, and the number of edge transfer network layers is set for the process graph structure.

3. The method for preparing a multi-node coordinated multi-layer composite thin film material according to claim 2, characterized in that: Fitting the change attenuation amount according to the dependency strength to determine the segmentation threshold includes: Determining the discreteness of the variation according to the dependency strength; Determining a center point according to the dispersion; According to the half amount of the change of the maximum dependence intensity and the second largest dependence intensity, the adjustment constant is obtained; The segmentation threshold is obtained according to the dependency strength corresponding to the central point and the adjustment constant.

4. The method for preparing a multi-node coordinated multi-layer composite thin film material according to claim 3, characterized in that: The expression of the segmentation threshold is: ,in, is the dependence strength corresponding to the center point, is the adjustment constant, ,in, is the maximum dependency strength, This is the highest level of dependence.

5. The method for preparing a multi-node coordinated multi-layer composite thin film material according to claim 2, characterized in that: The received feature information is aggregated, and each node uses an update function to combine the aggregated feature information with its own features to generate new node features, including: Set the influence weight of the edge transfer network layer; According to the influence weight, characteristic information is sequentially transmitted and accumulated downward from the most upstream node according to the hierarchical relationship of the process flow, and is transmitted to the downstream node through multiple aggregations to obtain aggregated characteristic information; By updating the function: , combining the aggregate feature information with its own features to obtain the new node features; wherein, is the new feature vector of node j at time t+1, is the activation function, is the influence weight of the upstream node on the downstream node, The set of upstream nodes at the edge transfer network level, is the feature of upstream node i at time t, is the feature vector of the current node j at time t, It is a trainable weight matrix used to perform linear transformation on the aggregated features.

6. The method for preparing a multi-node coordinated multi-layer composite thin film material according to claim 1, characterized in that: The step of performing preparation control parameter analysis according to the new node characteristics to obtain the node preparation parameters includes: Constructing a mapping function, wherein the mapping function is obtained by learning a conversion relationship using a historical training data set through a machine learning model, wherein the historical training data set includes a feature vector, a preparation control parameter and a mapping relationship thereof, wherein the mapping function is used to convert the feature vector into the preparation control parameter; The new node feature is used as an input variable, feature mapping is performed through the mapping function, and the preparation control parameter is obtained as the node preparation parameter.

7. The method for preparing a multi-node coordinated multi-layer composite thin film material according to claim 6, characterized in that: Obtaining the preparation control parameter as the node preparation parameter also includes: Set the equipment control parameter limit of each process node, use the equipment control parameter limit to determine whether the node preparation parameter meets the limit requirement, and implement hard constraints when it does not meet the limit to assign node preparation parameter constraints to the limit; Determine the deviation control parameter according to the node preparation parameter and the equipment control parameter limit, perform compensation analysis on the deviation control parameter according to the compensation parameter, and add the compensation parameter and the compensation control amount; The compensation parameters and the compensation control amount are integrated into the node preparation parameters to update the node preparation parameters.

8. A multi-node coordinated multi-layer composite thin film material preparation device, characterized in that: The device is used to implement the method for preparing a multi-node coordinated multi-layer composite thin film material according to any one of claims 1 to 7, and the device comprises: A process graph structure building module, wherein the process graph structure building module is used to construct a process graph structure by taking the preparation process flow of the multilayer composite film as a graph node, each process flow node as a vertex, and the association relationship between the nodes as an edge, wherein the association relationship is a physical parameter connection between the process flows; A constraint fitting module, the constraint fitting module is used to set timing constraints according to the timing relationship of the manufacturing process flow, fit the timing constraints to the edges, and the timing constraints are used to constrain the delay time from each node to the downstream node; A node feature generation module, wherein the node feature generation module is used to determine the process sequence according to the timing constraint condition, and each node transmits feature information of an adjacent upstream node through the edge in the process graph structure, and aggregates the received feature information, and each node uses an update function to combine the aggregated feature information with its own feature to generate a new node feature; The node preparation parameter acquisition module is used to perform preparation control parameter analysis according to new node characteristics to obtain node preparation parameters.

Images