Optimized control method and system for regeneration process of optoelectronic display material

Abstract

The application relates to the technical field of material science, and provides a photoelectric display material regeneration process optimization control method and system.The method comprises the following steps: acquiring a process node set, traversing a historical process node state vector, and collecting a historical extraction, recrystallization, purification and drying state vector set; traversing the vector set, performing associated mapping scoring on state variables and action variables according to node improvement variables, and constructing a hierarchical optimization strategy library; acquiring process nodes and state variables, combining the hierarchical optimization strategy library to perform process optimization control, determining an optimization scheme, and performing optimization control on the process nodes based on the scheme, so that the technical problems that process parameters of each key node in a traditional photoelectric display material regeneration process are complex and unstable, the regeneration process efficiency is low, and energy consumption is high are solved, the technical effect that the hierarchical optimization strategy library is used to realize accurate optimization of process parameters of key nodes, improve the recovery efficiency of the photoelectric display material regeneration process, and reduce energy consumption is achieved.

Description

Technical Field

[0001] This application relates to the field of materials science and technology, specifically to methods and systems for optimizing and controlling the recycling process of optoelectronic display materials. Background Technology

[0002] With the continuous development of display technology, optoelectronic display materials are widely used in liquid crystal displays (LCDs), organic light-emitting diodes (OLEDs), and other fields, becoming an indispensable part of modern electronic products. Especially driven by OLED technology, the demand for optoelectronic display materials is growing rapidly. However, the production process of optoelectronic display materials is complex, and the materials used typically include rare metals and organic compounds. These materials may generate residues, waste, or defective products during manufacturing. Therefore, how to effectively recycle, regenerate, and optimize these optoelectronic display materials has become an important issue for improving production efficiency, reducing production costs, saving energy and reducing emissions, and protecting the environment. However, existing optoelectronic display material regeneration processes typically face several challenges. First, optoelectronic display materials are easily affected by various factors during use, such as fluctuations in process parameters and changes in material properties, leading to low regeneration efficiency and low recovery rates. Second, traditional recycling processes often lack precise control over different materials and process nodes, resulting in excessive energy consumption and significant material loss during recycling. Furthermore, existing optimization control methods are mostly empirical adjustments, lacking intelligent optimization strategies and adaptive control methods. Summary of the Invention

[0003] This application provides an optimized control method and system for the recycling process of optoelectronic display materials, aiming to solve the technical problems of low efficiency and high energy consumption caused by the complex and unstable process parameters of key nodes in the traditional optoelectronic display material recycling process.

[0004] In a first aspect, the application discloses a method for optimizing control of a regeneration process of optoelectronic display materials, which comprises: obtaining a set of key regeneration process nodes of target optoelectronic display materials, traversing the set of key regeneration process nodes to collect historical regeneration key process node state vectors, and obtaining a set of historical extraction state vectors, a set of historical recrystallization state vectors, a set of historical purification state vectors and a set of historical drying state vectors, wherein each state vector comprises state variables, action variables and node index improvement variables; traversing the set of historical extraction state vectors, the set of historical recrystallization state vectors, the set of historical purification state vectors and the set of historical drying state vectors, performing associated mapping scoring on the state variables and the action variables according to the node index improvement variables, and constructing a hierarchical optimization strategy library according to the scoring results; obtaining a target process node and target process node state variables, combining the hierarchical optimization strategy library to perform regeneration process optimization control, determining a target regeneration process optimization control scheme, and performing optimization control on the target process node based on the target regeneration process optimization control scheme.

[0005] In another aspect, the application discloses a system for optimizing control of a regeneration process of optoelectronic display materials, which comprises: a historical data acquisition module: obtaining a set of key regeneration process nodes of target optoelectronic display materials, traversing the set of key regeneration process nodes to collect historical regeneration key process node state vectors, and obtaining a set of historical extraction state vectors, a set of historical recrystallization state vectors, a set of historical purification state vectors and a set of historical drying state vectors, wherein each state vector comprises state variables, action variables and node index improvement variables; an associated mapping module: traversing the set of historical extraction state vectors, the set of historical recrystallization state vectors, the set of historical purification state vectors and the set of historical drying state vectors, performing associated mapping scoring on the state variables and the action variables according to the node index improvement variables, and constructing a hierarchical optimization strategy library according to the scoring results; and an optimization control module: obtaining a target process node and target process node state variables, combining the hierarchical optimization strategy library to perform regeneration process optimization control, determining a target regeneration process optimization control scheme, and performing optimization control on the target process node based on the target regeneration process optimization control scheme.

[0006] The one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0007] The method for optimizing the regeneration process of the photoelectric display material first acquires key regeneration process nodes of a target material and collects relevant historical process data, which includes state information of multiple process stages (such as extraction, recrystallization, purification, and drying), each state containing multiple variables, such as state variables, action variables, and node improvement indicators. Subsequently, the historical data is analyzed, and the state variables and action variables are scored through node indicators, thereby constructing a hierarchical optimization strategy library that provides optimization suggestions for each process node. Finally, by acquiring the target process node and its state variables, the optimization strategy library is combined to perform regeneration process optimization control, thereby formulating a specific optimization scheme and adjusting the process node according to the scheme to improve the efficiency, recovery rate, and product quality of the regeneration process.

[0008] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the specific embodiments of the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0010] Figure 1 A flowchart of the photoelectric display material regeneration process optimization control method in an embodiment.

[0011] Figure 2 A system architecture diagram of the photoelectric display material regeneration process optimization control system in an embodiment.

[0012] Legend: historical data acquisition module 11, association mapping module 12, optimization control module 13. DETAILED DESCRIPTION

[0013] The embodiments of the present application provide a photoelectric display material regeneration process optimization control method and system, which solves the technical problem of complex and unstable process parameters of each key node in the traditional photoelectric display material regeneration process, resulting in low regeneration process efficiency and high energy consumption.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such process, method, product, or device.

[0016] Example 1, as Figure 1 As shown, this application provides a method for optimizing and controlling the recycling process of optoelectronic display materials, the method comprising:

[0017] Obtain the set of key regeneration process nodes for the target optoelectronic display material, traverse the set of key regeneration process nodes to collect historical regeneration key process node state vectors, and obtain the set of historical extraction state vectors, the set of historical recrystallization state vectors, the set of historical purification state vectors, and the set of historical drying state vectors. Each state vector includes state variables, action variables, and node index improvement variables.

[0018] In the embodiment of the present application, in the process of optimizing the regeneration process of optoelectronic display materials, first, the key regeneration process node set of the target material is obtained, which includes various regeneration links of the material, such as extraction, recrystallization, purification and drying, etc. Each link has its unique operation requirements and parameters, for example, the extraction link involves parameters such as solvent selection, temperature control, stirring rate, etc., which are used to separate the target components in the waste material; the recrystallization link involves parameters such as crystallization temperature, solvent ratio, cooling rate, etc., which are used to restore the crystal structure of the material. By collecting and analyzing the historical data of these nodes, sufficient basis can be provided for optimization control. Subsequently, the key regeneration process node set is traversed, and for each traversed node, the historical regeneration key process node state vector corresponding to the node is collected from the historical regeneration process log to form a corresponding set, such as the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set. Each state vector in the set is composed of state variables, action variables and node index improvement variables, wherein the state variables describe the state information of the current process node, including the properties of the raw materials (such as impurity concentration, viscosity, pH value, optical purity, etc.) and equipment parameters (such as reaction kettle temperature, stirring speed, pressure, etc.), reflecting the physical and chemical environment of the current process node, which helps to understand the performance of the material in the current process link. The action variable represents the operation instruction taken at a specific time, which usually includes operation parameters directly related to the process, such as heating rate, solvent injection flow, stirring intensity, cooling rate, etc., which can ensure that the process effect reaches the expected target. The node index improvement variable represents the improvement value of the target index, which is used to evaluate the operation effect of the process node, usually including purity improvement percentage, energy consumption reduction value, product yield change, etc., reflecting the improvement degree of the quality and process performance of the target optoelectronic display material after operation, which provides data basis for the construction of subsequent correlation mapping score and hierarchical optimization strategy library.

[0019] Traversing the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set, the state variables and action variables are correlated and mapped according to the node index improvement variable, and a hierarchical optimization strategy library is constructed according to the scoring results.

[0020] In one embodiment, after obtaining the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set, the sets are traversed, and for each traversed state vector, the node index improvement variable, the state variable and the action variable in the state vector are associated and mapped by using the pre-constructed association mapping scorer, that is, the association mapping scorer uses the node index improvement variable to analyze the state variable and the action variable according to the learned mapping relationship, determines the action mapping score corresponding to the historical state, and then optimizes the corresponding historical state vector set according to the action mapping scores to determine a plurality of hierarchical optimization sub-policy libraries, for example, the action mapping scores of the historical extraction state are used to hierarchically optimize the historical extraction state vector set, and the determined hierarchical optimization sub-policy library is an extraction hierarchical optimization sub-policy library. For each hierarchical optimization sub-policy library, it includes high-level strategies and low-level strategies, wherein the high-level strategies are used to make decisions in a more abstract state space, select the skills that should be used at present, such as reinforced extraction, energy-saving drying, etc., and represent how to handle different states under different process nodes; the low-level strategies are more specific parameter variables, which are responsible for executing operation instructions in the original action space according to the skills selected by the high-level strategies, and these operation instructions are usually action variables that directly affect the process effect, such as heating rate, solvent flow, stirring intensity, etc. By connecting all the obtained hierarchical optimization sub-policy libraries in parallel, a final hierarchical optimization strategy library is formed, which contains optimization strategies at different levels to ensure accurate control and adjustment according to different actual situations in complex process, thereby maximizing the regeneration effect and production efficiency of the optoelectronic display material.

[0021] Further, the application provides traversing the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set, associating and mapping the state variable and the action variable according to the node index improvement variable, and constructing a hierarchical optimization strategy library according to the scoring results, comprising:

[0022] The pre-constructed association mapping scorer respectively performs association mapping scoring on the state variables and action variables in the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set according to the node index improvement variables, determines a historical extraction state-action mapping score set, a historical recrystallization state-action mapping score set, a historical purification state-action mapping score set and a historical drying state-action mapping score set; performs hierarchical optimization analysis according to the historical extraction state vector set and the historical extraction state-action mapping score set, and constructs an extraction hierarchical optimization sub-strategy library; performs hierarchical optimization analysis according to the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set, and the historical recrystallization state-action mapping score set, the historical purification state-action mapping score set and the historical drying state-action mapping score set, and constructs a purification hierarchical optimization sub-strategy library, a recrystallization hierarchical optimization sub-strategy library and a drying hierarchical optimization sub-strategy library; and the extraction hierarchical optimization sub-strategy library, the purification hierarchical optimization sub-strategy library, the recrystallization hierarchical optimization sub-strategy library and the drying hierarchical optimization sub-strategy library are connected in parallel to construct the hierarchical optimization strategy library.

[0023] Preferably, before the association mapping score, the sample state vector (including sample state variables, sample action variables and sample node index improvement variables) and the sample action mapping score are divided into a training set and a validation set, and the initial association mapping scorer is trained through the training set, and after the training is completed, the performance of the model is evaluated using the validation set, wherein the initial association mapping scorer can be constructed based on a fully connected neural network, a multi-layer perceptron, a deep neural network, etc. Taking a fully connected neural network as an example, an initial association mapping scorer structure is constructed using a fully connected neural network, including an input layer, an output layer, a hidden layer, etc. The weights of the initial association mapping scorer are initialized using random numbers, and the training set is input into the initialized initial association mapping scorer for forward propagation, which is transmitted layer by layer through the input layer, the hidden layer and the output layer, and the mapping result including the action mapping score is calculated. Subsequently, the loss value between the mapping result and the sample data is calculated using the mean square error loss function, and the gradient of the loss to the weights of each layer is calculated through back propagation, and the Adam optimizer is used to optimize the scorer parameters to adjust the weights to minimize the value of the loss function. The above process is repeated until the maximum number of iterations is reached. After the training is completed, the performance of the scorer is tested using the validation set to evaluate the accuracy of the scorer in the action mapping score mapping task. If the accuracy meets the expected accuracy, the current initial association mapping scorer is output as the final association mapping scorer, otherwise, the learning rate, the number of training batches and other hyperparameters are adjusted to further improve the mapping effect of the association mapping scorer. After constructing the association mapping scorer corresponding to each link, the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set are input into the corresponding association mapping scorer. The association mapping scorer analyzes the state variables and action variables based on the received node index improvement variables, and calculates multiple action mapping scores. By associating these action mapping scores with the corresponding historical states (such as historical extraction states, historical recrystallization states, etc.), the historical extraction state-action mapping score set, the historical recrystallization state-action mapping score set, the historical purification state-action mapping score set and the historical drying state-action mapping score set are organized. Then, the similarity of each vector in the historical extraction state vector set is used to divide the calculated historical extraction state-action mapping score set, and the hierarchical extraction state vector set is extracted according to the division result to determine the high-level extraction state vector and the low-level action variable, and a extraction hierarchical optimization sub-strategy library is constructed based on these vectors and variables.For the historical recrystallization state vector set and the historical recrystallization state-action mapping score set, the historical purification state vector set and the historical purification state-action mapping score set, and the historical drying state vector set and the historical drying state-action mapping score set, the same hierarchical optimization analysis as described above is performed to construct a purification hierarchical optimization sub-strategy library, a recrystallization hierarchical optimization sub-strategy library, and a drying hierarchical optimization sub-strategy library. Finally, the extraction hierarchical optimization sub-strategy library, the purification hierarchical optimization sub-strategy library, the recrystallization hierarchical optimization sub-strategy library, and the drying hierarchical optimization sub-strategy library are connected in parallel to form a complete hierarchical optimization strategy library. This hierarchical optimization strategy library can efficiently coordinate the optimization process between different process nodes and ensure that the control of the entire regeneration process can be accurately adjusted according to the actual situation.

[0024] Further, the application provides hierarchical optimization analysis based on the historical extraction state vector set and the historical extraction state-action mapping score set to construct an extraction hierarchical optimization sub-strategy library, including:

[0025] Based on the state vector similarity, the historical extraction state vector set is divided into M divided historical extraction state vector sets, and the historical extraction state-action mapping score set is divided into M divided historical extraction state-action mapping score sets according to the M divided historical extraction state vector sets, wherein M is a positive integer. High-level extraction state vectors are determined by high-level extraction of the M divided historical extraction state vector sets. From the dimensions of score size and action variable aggregation degree, the state variables in the M divided historical extraction state vector sets are screened in combination with the M divided historical extraction state-action mapping score sets to determine M low-level action variables. The M high-level extraction state vectors and M low-level action variables are associated one-to-one to construct the extraction hierarchical optimization sub-strategy library.

[0026] Optionally, in the hierarchical optimization of the extraction process, first, the relationship between the intra-cluster error sum of squares and the number of clusters is plotted, and the position where the error reduction speed slows down significantly is selected as the initial clustering number M (M is an integer), and then M historical extraction state vectors are randomly selected from the historical extraction state vector set as cluster centers. Subsequently, for each historical extraction state vector, the cosine similarity is used to calculate the similarity between the historical extraction state vector and all cluster center state vectors, and it is assigned to the cluster with the highest state vector similarity. After the assignment is completed, the mean of all state vectors in each cluster is calculated as the new cluster center, and the above process is repeated until the cluster center no longer changes or the change is small enough, indicating that the clustering process has converged, at this time, M partitioned historical extraction state vector sets are obtained. Then, according to the historical extraction state vectors in each partitioned historical extraction state vector set, the corresponding action mapping score in the historical extraction state-action mapping score set is extracted, and is aggregated according to the clustering cluster to which the historical extraction state vector belongs, and M partitioned historical extraction state-action mapping score sets are constructed. Further, the mean vector of each partitioned historical extraction state vector set is calculated, and the representative state vector is extracted through iteration according to the average state vector, thereby obtaining M high-level extraction state vectors of the M partitioned historical extraction state vector sets. Then, for each partitioned historical extraction state vector set, the aggregation degree of each action variable is calculated, the low aggregation degree of the action variable indicates a larger selection range in that state, and the high aggregation degree indicates a consistent action selection trend, and the calculated aggregation degree is combined with the partitioned historical extraction state-action mapping score set to filter out the action variable that best represents the process control in that state, to form M low-level action variables. Generally, action variables with higher scores and lower aggregation degrees are selected because these variables may have greater optimization potential. Finally, the extracted M high-level extraction state vectors and M low-level action variables are associated one-to-one to form a set of state-action pairs, and by adding these state-action pairs one by one to a set, an extraction hierarchical optimization sub-strategy library is formed, which can help the system better perform dynamic optimization and decision-making, thereby achieving a more efficient and energy-saving production process.

[0027] Further, the application provides high-level extraction of the M partitioned historical extraction state vector sets to determine M high-level extraction state vectors, comprising:

[0028] The mean of the M partitioned historical extraction state vector sets is calculated to determine M standard condition historical extraction state vectors; the M standard condition historical extraction state vectors are extracted in a random direction in the M partitioned historical extraction state vector sets according to a preset extraction vector tolerance bandwidth to determine M iteration historical extraction state vectors; an update iteration direction is determined based on the M iteration historical extraction state vectors and the M standard condition historical extraction state vectors, and high-level extraction is performed based on the update iteration direction to determine M high-level extraction state vectors.

[0029] Optionally, for the M partitioned historical extraction state vector sets, the mean of all historical extraction state vectors in each partitioned historical extraction state vector set is calculated to obtain a standard condition historical extraction state vector of each partitioned historical extraction state vector set. Subsequently, an extraction vector tolerance bandwidth preset according to business requirements is obtained, and the extraction vector tolerance bandwidth defines the range of deviation allowed by the state vector in each extraction. For each standard condition historical extraction state vector, a small random disturbance in each direction is performed according to the preset extraction vector tolerance bandwidth, that is, the current standard condition historical extraction state vector is added to the product of the disturbance amplitude and the random number, thereby obtaining M iteration historical extraction state vectors, wherein the disturbance amplitude is determined by the preset extraction vector tolerance bandwidth, and the random number is uniformly distributed data in each dimension. Then, the difference between each standard condition historical extraction state vector and the iteration historical extraction state vector is calculated, and the update direction is adjusted according to the difference to obtain an update iteration direction. Then, the value of the standard condition historical extraction state vector is adjusted according to the update iteration direction. After several iterations, the obtained state vector is a high-level extraction state vector. These high-level extraction state vectors represent the key states in the process and can maximize the overall characteristics of the process node, thereby providing an important decision basis for subsequent process optimization control.

[0030] Further, the application provides that the M iteration historical extraction state vectors and the M standard condition historical extraction state vectors are used to determine an update iteration direction, high-level extraction is performed based on the update iteration direction, and M high-level extraction state vectors are determined, which includes:

[0031] The cosine similarity calculation formula is used to calculate the similarity between each of the M iteration historical extraction state vectors and the historical extraction state vectors in the corresponding divided historical extraction state vector set, the number of calculation results greater than or equal to a preset similarity threshold is counted, and M iteration historical extraction state vector update coefficients are obtained. The cosine similarity calculation formula is used to calculate the similarity between each of the M standard historical extraction state vectors and the historical extraction state vectors in the corresponding divided historical extraction state vector set, the number of calculation results greater than or equal to a preset similarity threshold is counted, and M standard historical extraction state vector update coefficients are obtained. When the M iteration historical extraction state vector update coefficients are less than the M standard historical extraction state vector update coefficients, the direction from the M iteration historical extraction state vectors to the M standard historical extraction state vectors is taken as an update iteration direction, the M standard historical extraction state vectors are iterated according to the preset extraction vector tolerance bandwidth in the update iteration direction, until a preset iteration number is met, and M high-level extraction state vectors are obtained. When the M iteration historical extraction state vector update coefficients are greater than or equal to the M standard historical extraction state vector update coefficients, the direction from the M standard historical extraction state vectors to the M iteration historical extraction state vectors is taken as an update iteration direction, the M iteration historical extraction state vectors are iterated according to the preset extraction vector tolerance bandwidth in the update iteration direction, until a preset iteration number is met, and M high-level extraction state vectors are obtained.

[0032] Optionally, after obtaining the M iteration history extracted state vectors, the cosine similarity is used to calculate the similarity between each iteration history extracted state vector and the historical extracted state vector in the corresponding partitioned historical extracted state vector set, and the calculated multiple similarities are compared with the preset similarity threshold. The number of similarities greater than or equal to the preset similarity threshold is counted, wherein the preset similarity threshold is the minimum similarity size set by the person skilled in the art to meet the requirements. The number is the update coefficient of the iteration history extracted state vector. The larger the number is, the more similar the historical extracted state vectors to the iteration history extracted state vector are, and the higher the importance is. Similarly, the update coefficient of each landmark history extracted state vector is also calculated according to the M partitioned historical extracted state vector set. Then, for each pair of landmark history extracted state vector and iteration history extracted state vector, the update coefficients thereof are compared. If the update coefficient of the iteration history extracted state vector is less than that of the landmark history extracted state vector, it indicates that the landmark history extracted state vector is closer to the historical state vector. Therefore, the update is performed in the direction from the iteration history extracted state vector to the landmark history extracted state vector. If the update coefficient of the iteration history extracted state vector is greater than or equal to that of the landmark history extracted state vector, it indicates that the iteration history extracted state vector is closer to the historical state vector. The update should be performed in the direction from the landmark history extracted state vector to the iteration history extracted state vector. If the direction from the iteration state vector to the landmark state vector is selected, the iteration history extracted state vector is continuously adjusted in this direction using the preset extracted vector tolerance bandwidth. If the direction from the landmark state vector to the iteration state vector is selected, the iteration history extracted state vector is continuously adjusted in this direction. This process continues until the preset iteration number is reached or the state vector variation amplitude is less than the threshold. After the iteration ends, the current M iteration history extracted state vectors are taken as the M high-level extracted state vectors, which can represent the characteristics of the partition set and provide important decision basis for subsequent optimization control.

[0033] Further, the application provides screening of state variables in the M partitioned historical extracted state vector set from two dimensions of score size and action variable aggregation degree, in combination with the M partitioned historical extracted state-action mapping score set, to determine M low-level action variables, comprising:

[0034] According to the preset aggregation bandwidth, the density of action variables aggregated around each action variable in the M partitioned historical extracted state vector set is counted to determine the M partitioned action variable aggregation degree set. The state variables in the M partitioned historical extracted state vector set are screened from two dimensions of score size and action variable aggregation degree, in combination with the M partitioned historical extracted state-action mapping score set and the M partitioned action variable aggregation degree set, to determine M low-level action variables.

[0035] Optionally, to achieve filtering based on both rating size and action variable clustering, a preset clustering bandwidth is first obtained. This preset clustering bandwidth determines the range of action variable clustering. For each action variable in the M partitioned historical extracted state vector sets, Euclidean distance is used to calculate the distance between each action variable and other action variables. The calculated distance is compared with the preset clustering bandwidth. If the distance falls within the preset clustering bandwidth, it indicates that the two action variables have a clustering relationship. By counting the number of action variables that meet the preset clustering bandwidth requirement and then dividing the number of action variables that meet the requirement by the total number of action variables, the density of action variables clustered around each action variable can be obtained. These action variable densities are stored in the order of the M partitioned historical extracted state vector sets, forming the M partitioned historical extracted state vector sets. Subsequently, state variables in the M sets of historical extracted state vectors are filtered based on two dimensions: score magnitude and action variable clustering. During this process, the action mapping score for each historical extracted state vector is obtained from the M sets of historical extracted state-action mapping score sets. Then, the action mapping score of each historical extracted state vector is weighted and calculated with the action variable density in the corresponding action variable clustering set to obtain a comprehensive score for each action variable. A higher comprehensive score indicates a more important action variable. Next, the action variables are sorted in descending order based on the calculated comprehensive scores, and the sorted results are segmented according to a score segmentation threshold. Segments with scores greater than or equal to the score segmentation threshold are summarized and organized into M low-level action variables. These low-level action variables represent the optimal operational instructions in a given state, and they play a crucial role in subsequent optimization processes, ensuring that the process operation achieves optimal results.

[0036] Obtain the target process node and its state variables, combine the hierarchical optimization strategy library to perform regeneration process optimization control, determine the target regeneration process optimization control scheme, and optimize the target process node based on the target regeneration process optimization control scheme.

[0037] In one embodiment, after the hierarchical optimization strategy library is constructed, the target process node and its corresponding target process node state variable are obtained to provide a basis for subsequent optimization control, where the target process node generally represents a key link in the entire regeneration process, such as extraction, recrystallization, purification or drying, etc.; the target process node state variable reflects the current conditions of process operation, such as impurity concentration, viscosity, pH value, reaction kettle temperature, etc. Subsequently, according to the target process node and the target process node state variable, matching is performed in the hierarchical optimization strategy library, the most suitable strategy is selected, and a target regeneration process optimization control scheme is generated according to the strategy, which includes a series of operation parameters and optimization objectives, such as how to adjust the heating temperature, control the solvent flow, or change the stirring intensity under a certain state, etc. Finally, based on the determined target regeneration process optimization control scheme, specific optimization control is performed for the target process node, and the optimization control process includes adjusting the process operation parameters according to the control scheme and monitoring the process execution state in real time to ensure that the process always runs under optimized conditions. When the state of the process node changes, the control parameters are dynamically adjusted to adapt to the new process requirements, so as to continuously maintain the efficiency and stability of the entire regeneration process, improve the material recovery rate, and reduce energy consumption.

[0038] Further, the application provides obtaining the target process node and the target process node state variable, and performing regeneration process optimization control in combination with the hierarchical optimization strategy library to determine the target regeneration process optimization control scheme, including:

[0039] Based on the target process node and the target process node state variable, the strategy matching is performed on the hierarchical optimization strategy library to determine the matching action variable; the matching action variable is adjusted multiple times in action amplitude to determine the adjusted action variable set; the updated state-action correlation reliability prediction branch is obtained, and the target process node state variable and the adjusted action variable set are predicted in reliability by using the updated state-action correlation reliability prediction branch to determine the adjusted reliability set; the adjusted action variable corresponding to the maximum value in the adjusted reliability set is taken as the adjustment direction, and the adjusted action variables corresponding to the remaining adjusted reliabilities are adjusted according to the preset adjustment amplitude to determine the following adjusted action variable set and the following adjusted reliability set; when there is an adjusted reliability greater than the adjusted reliability corresponding to the adjustment direction in the following adjusted reliability set, the adjusted action variable corresponding to the maximum value in the following adjusted reliability set is taken as the updated adjustment direction, and the remaining following adjusted action variable set is continuously adjusted until the preset adjustment times are met, the target adjusted action variable is obtained, and the target adjusted action variable is taken as the target regeneration process optimization control scheme.

[0040] Preferably, the target process node and the target process node state variable are first matched with the process nodes and state variables in the hierarchical optimization strategy library, the similarity of the target process node state variable and the state variables in the hierarchical optimization strategy library is calculated by cosine similarity, and the state vector corresponding to the maximum similarity is extracted as the matching action variable. These matching action variables will be used for subsequent adjustment and optimization. Subsequently, a plurality of possible operation combinations are obtained by randomly adjusting the matching action variables within the allowable range of device parameters, and these operation combinations are summarized as an adjustment action variable set. Then, the state-action associated reliability prediction branch is used to predict the reliability of the target process node state variable and the plurality of operation combinations in the adjustment action variable set, and the adjustment reliability of each operation combination is obtained. By summarizing these adjustment reliabilities, an adjustment reliability set is formed. Then, in the adjustment reliability set, the adjustment action variable of the operation combination with the highest adjustment reliability is taken as the adjustment direction, and the remaining adjustment action variables are adjusted according to the preset adjustment amplitude. This preset adjustment amplitude is a fixed value. After adjustment, a new following adjustment action variable set is determined, and the state-action associated reliability prediction branch is used to analyze the following adjustment action variable set to determine the following adjustment reliability set. If it is found in the following adjustment reliability set that the reliability value of a certain adjustment action variable is greater than the reliability corresponding to the adjustment direction, it means that this action variable has the potential to better optimize the process, and therefore this action variable is taken as the new adjustment direction. After updating the adjustment direction, the remaining following adjustment action variables are adjusted until the preset adjustment times are met or until the reliability difference tends to be stable. The reliability value and the action variable after each adjustment are updated, and whether further adjustment is needed is determined according to the new reliability. After multiple rounds of adjustment and optimization, a set of target adjustment action variables are finally obtained, which can achieve optimal process control under the current process conditions and ensure that the photoelectric display material regeneration process can be efficiently and stably performed in actual production.

[0041] Further, when there is no adjustment reliability greater than the adjustment reliability corresponding to the adjustment direction in the following adjustment reliability set, the following adjustment action variable set is adjusted in the adjustment direction according to the preset adjustment amplitude until the preset adjustment times are met, and a target adjustment action variable is obtained. The target adjustment action variable is taken as the target regeneration process optimization control scheme.

[0042] Optionally, if no adjustment reliability greater than the adjustment reliability corresponding to the adjustment direction is followed, the adjustment reliability set is adjusted according to the preset adjustment amplitude in the adjustment direction, and the adjustment process is continued until the preset number of adjustments is met. After a series of adjustments, a set of target adjustment action variables are finally determined, which will become the core of the optimization control scheme. Through these target adjustment action variables, it can be ensured that the regeneration process of the optoelectronic display material is always in the optimal operating state, and the stability and efficiency of the regeneration process of the optoelectronic display material are ensured.

[0043] Further, the application provides an updated state-action associated reliability prediction branch, comprising:

[0044] An initial state-action associated reliability prediction branch is obtained, and the initial state-action associated reliability prediction branch is adjusted in parameters by using the node index improvement variables and action variables in the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set to obtain an updated state-action associated reliability prediction branch after parameter updating.

[0045] Optionally, first, a fully connected neural network is designed and initialized as an initial state-action associated reliability prediction branch, and then the initial state-action associated reliability prediction branch is adjusted in parameters according to the node index improvement variables and action variables in the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set and the marked adjustment reliability, that is, the weights and bias terms in the initial state-action associated reliability prediction branch are adjusted through the foregoing similar forward propagation, loss calculation, back propagation and parameter optimization steps, so that the prediction ability of the initial state-action associated reliability prediction branch is continuously enhanced, and the state variables and action variables can be more accurately mapped to the adjustment reliability prediction value. After completing the iterative training, the current initial state-action associated reliability prediction branch is used as an updated state-action associated reliability prediction branch, which can provide more accurate state-action mapping prediction and accurately provide optimization control for the target process node, so as to ensure that the optimization control scheme can be dynamically adjusted according to the actual feedback, and improve the process efficiency and quality.

[0046] In summary, the embodiments of the application have at least the following technical effects:

[0047] The embodiment of the application first acquires a key regeneration process node set of a target optoelectronic display material, traverses the key regeneration process node set to collect a historical regeneration key process node state vector, and obtains a historical extraction state vector set, a historical recrystallization state vector set, a historical purification state vector set and a historical drying state vector set, wherein each state vector includes a state variable, an action variable and a node index improvement variable; then, the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set are traversed, the state variable and the action variable are associated and mapped according to the node index improvement variable, and a hierarchical optimization strategy library is constructed according to the scoring result; finally, a target process node and a target process node state variable are acquired, the hierarchical optimization strategy library is combined to perform regeneration process optimization control, a target regeneration process optimization control scheme is determined, and the target process node is optimized and controlled based on the target regeneration process optimization control scheme. These technical effects jointly solve the technical problems of complex and unstable process parameters of each key node in the traditional optoelectronic display material regeneration process, resulting in low regeneration process efficiency and high energy consumption, and achieve the technical effects of realizing accurate optimization of key node process parameters through a hierarchical optimization strategy library, improving the recovery efficiency of the optoelectronic display material regeneration process, and reducing energy consumption.

[0048] In the embodiment two, based on the same inventive concept as the optoelectronic display material regeneration process optimization control method in the foregoing embodiments, as shown in the embodiment two, the application provides an optoelectronic display material regeneration process optimization control system, which includes: Figure 2 The system includes: a historical data acquisition module 11: acquiring a key regeneration process node set of a target optoelectronic display material, traversing the key regeneration process node set to collect a historical regeneration key process node state vector, and obtaining a historical extraction state vector set, a historical recrystallization state vector set, a historical purification state vector set and a historical drying state vector set, wherein each state vector includes a state variable, an action variable and a node index improvement variable; an associated mapping module 12: traversing the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set, associated mapping and scoring the state variable and the action variable according to the node index improvement variable, and constructing a hierarchical optimization strategy library according to the scoring result; an optimization control module 13: acquiring a target process node and a target process node state variable, combining the hierarchical optimization strategy library to perform regeneration process optimization control, determining a target regeneration process optimization control scheme, and optimizing and controlling the target process node based on the target regeneration process optimization control scheme.

[0049] Further, the associated mapping module 12 is further used to perform the following method:

[0050] The pre-constructed correlation mapping scorer respectively performs correlation mapping scoring on the state variables and action variables in the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set according to the node index improvement variable, to determine a historical extraction state-action mapping score set, a historical recrystallization state-action mapping score set, a historical purification state-action mapping score set and a historical drying state-action mapping score set; hierarchical optimization analysis is performed according to the historical extraction state vector set and the historical extraction state-action mapping score set, to construct an extraction hierarchical optimization sub-strategy library; hierarchical optimization analysis is performed according to the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set, and the historical recrystallization state-action mapping score set, the historical purification state-action mapping score set and the historical drying state-action mapping score set, to construct a purification hierarchical optimization sub-strategy library, a recrystallization hierarchical optimization sub-strategy library and a drying hierarchical optimization sub-strategy library; the extraction hierarchical optimization sub-strategy library, the purification hierarchical optimization sub-strategy library, the recrystallization hierarchical optimization sub-strategy library and the drying hierarchical optimization sub-strategy library are connected in parallel to construct the hierarchical optimization strategy library.

[0051] Further, the correlation mapping module 12 is further used to perform the following method:

[0052] Based on the state vector similarity, the historical extraction state vector set is divided into M divided historical extraction state vector sets, and the historical extraction state-action mapping score set is divided into M divided historical extraction state-action mapping score sets according to the M divided historical extraction state vector sets, where M is a positive integer; high-level extraction is performed on the M divided historical extraction state vector sets to determine M high-level extraction state vectors; from the dimensions of score size and action variable aggregation degree, the state variables in the M divided historical extraction state vector sets are screened in combination with the M divided historical extraction state-action mapping score sets to determine M low-level action variables; the M high-level extraction state vectors and the M low-level action variables are associated one by one to construct the extraction hierarchical optimization sub-strategy library.

[0053] Further, the correlation mapping module 12 is further used to perform the following method:

[0054] calculate the mean of the M partition history extraction state vector sets to determine M standard condition history extraction state vectors; perform random direction extraction iteration on the M standard condition history extraction state vectors in the M partition history extraction state vector sets according to a preset extraction vector tolerance bandwidth to determine M iteration history extraction state vectors; determine an update iteration direction based on the M iteration history extraction state vectors and the M standard condition history extraction state vectors, and perform high-level extraction based on the update iteration direction to determine M high-level extraction state vectors.

[0055] Further, the association mapping module 12 is further configured to perform the following method:

[0056] Calculate the similarity of each iteration history extraction state vector in the M iteration history extraction state vectors and the history extraction state vector in the corresponding partition history extraction state vector set using the cosine similarity calculation formula, and count the number of calculation results greater than or equal to the preset similarity threshold to obtain M iteration history extraction state vector update coefficients; calculate the similarity of each standard condition history extraction state vector in the M standard condition history extraction state vectors and the history extraction state vector in the corresponding partition history extraction state vector set using the cosine similarity calculation formula, and count the number of calculation results greater than or equal to the preset similarity threshold to obtain M standard condition history extraction state vector update coefficients; when the M iteration history extraction state vector update coefficients are less than the M standard condition history extraction state vector update coefficients, the direction from the M iteration history extraction state vectors to the M standard condition history extraction state vectors is taken as the update iteration direction, and the M standard condition history extraction state vectors are iterated according to the preset extraction vector tolerance bandwidth according to the update iteration direction until the preset iteration number is met, and M high-level extraction state vectors are obtained; when the M iteration history extraction state vector update coefficients are greater than or equal to the M standard condition history extraction state vector update coefficients, the direction from the M standard condition history extraction state vectors to the M iteration history extraction state vectors is taken as the update iteration direction, and the M iteration history extraction state vectors are iterated according to the preset extraction vector tolerance bandwidth according to the update iteration direction until the preset iteration number is met, and M high-level extraction state vectors are obtained.

[0057] Further, the association mapping module 12 is further configured to perform the following method:

[0058] According to the preset aggregation bandwidth, the density of action variables aggregated around each action variable in the M partition history extraction state vector set is counted, and M partition action variable aggregation degree sets are determined; from the dimensions of score size and action variable aggregation degree, the state variables in the M partition history extraction state vector set are screened in combination with the M partition history extraction state-action mapping score set and the M partition action variable aggregation degree set, and M low-level action variables are determined.

[0059] Further, the optimization control module 13 is also used to execute the following method:

[0060] Based on the target process node and the target process node state variable, the layered optimization strategy library is matched with a strategy to determine a matching action variable; the matching action variable is adjusted in multiple action amplitudes to determine an adjusted action variable set; an updated state-action correlation reliability prediction branch is obtained, and the target process node state variable and the adjusted action variable set are respectively predicted in reliability using the updated state-action correlation reliability prediction branch to determine an adjusted reliability set; the adjusted action variable corresponding to the maximum value in the adjusted reliability set is taken as an adjustment direction, and the adjusted action variables corresponding to the remaining adjusted reliabilities are adjusted according to a preset adjustment amplitude to determine a following adjusted action variable set and a following adjusted reliability set; when there is an adjusted reliability greater than the adjusted reliability corresponding to the adjustment direction in the following adjusted reliability set, the adjusted action variable corresponding to the maximum value in the following adjusted reliability set is taken as an updated adjustment direction, and the remaining following adjusted action variables are continuously adjusted until a preset adjustment number is met, a target adjusted action variable is obtained, and the target adjusted action variable is taken as the target regenerative process optimization control scheme.

[0061] Further, the optimization control module 13 is also used to execute the following method:

[0062] When there is no adjusted reliability greater than the adjusted reliability corresponding to the adjustment direction in the following adjusted reliability set, the following adjusted action variable set is continuously adjusted according to a preset adjustment amplitude to the adjustment direction until a preset adjustment number is met, a target adjusted action variable is obtained, and the target adjusted action variable is taken as the target regenerative process optimization control scheme.

[0063] Further, the optimization control module 13 is also used to execute the following method:

[0064] Obtaining an initial state-action correlation reliability prediction branch; using the node index improvement variables and action variables in the historical extraction state vector set, the historical recrystallization state vector set, the historical purification state vector set and the historical drying state vector set to perform parameter update adjustment on the initial state-action correlation reliability prediction branch, and obtaining an updated state-action correlation reliability prediction branch after parameter update.

[0065] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0066] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0067] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that fall within the scope of the present application are considered to be covered by the present application. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is intended to include these modifications and variations.

Claims

1. A method for optimizing and controlling the recycling process of optoelectronic display materials, characterized in that, The method includes: Obtain the set of key regeneration process nodes for the target optoelectronic display material, traverse the set of key regeneration process nodes to collect historical regeneration key process node state vectors, and obtain the set of historical extraction state vectors, the set of historical recrystallization state vectors, the set of historical purification state vectors and the set of historical drying state vectors. Each state vector includes state variables, action variables and node index improvement variables. Traverse the historical extraction state vector set, historical recrystallization state vector set, historical purification state vector set, and historical drying state vector set; perform correlation mapping and scoring on the state variables and action variables based on the node indicator improvement variables; and construct a hierarchical optimization strategy library based on the scoring results. Obtain the target process node and its state variables, combine the hierarchical optimization strategy library to execute regeneration process optimization control, determine the target regeneration process optimization control scheme, and optimize the target process node based on the target regeneration process optimization control scheme. The system iterates through the historical extraction state vector set, historical recrystallization state vector set, historical purification state vector set, and historical drying state vector set. Based on the node indicator improvement variables, it performs correlation mapping and scoring on the state variables and action variables. Then, based on the scoring results, it constructs a hierarchical optimization strategy library, including: A pre-constructed correlation mapping scorer is used to perform correlation mapping scores on state variables and action variables in the historical extraction state vector set, historical recrystallization state vector set, historical purification state vector set, and historical drying state vector set, respectively, based on the node index improvement variables, to determine the historical extraction state-action mapping score set, historical recrystallization state-action mapping score set, historical purification state-action mapping score set, and historical drying state-action mapping score set. Based on the historical extraction state vector set and the historical extraction state-action mapping score set, a hierarchical optimization analysis is performed to construct an extraction hierarchical optimization sub-strategy library; Perform the same hierarchical optimization analysis as described above. Based on the historical recrystallization state vector set, historical purification state vector set, and historical drying state vector set, as well as the historical recrystallization state-action mapping score set, historical purification state-action mapping score set, and historical drying state-action mapping score set, construct a purification hierarchical optimization sub-strategy library, a recrystallization hierarchical optimization sub-strategy library, and a drying hierarchical optimization sub-strategy library. The extraction layering optimization sub-strategy library, purification layering optimization sub-strategy library, recrystallization layering optimization sub-strategy library, and drying layering optimization sub-strategy library are connected in parallel to construct the layering optimization strategy library; Based on the historical extraction state vector set and the historical extraction state-action mapping score set, a hierarchical optimization analysis is performed to construct an extraction hierarchical optimization sub-strategy library, including: Based on state vector similarity, the historical extracted state vector set is divided into states to determine M sets of historical extracted state vector sets. Then, based on the M sets of historical extracted state vector sets, the historical extracted state-action mapping score set is divided into M sets of historical extracted state-action mapping score sets, where M is a positive integer. High-level extraction is performed on the M sets of partitioned historical extraction state vectors to determine M high-level extraction state vectors; Based on two dimensions—score size and action variable clustering—the state variables in the M historical extracted state-action mapping score sets are filtered to determine the M low-level action variables. The M high-level extracted state vectors and M low-level action variables are associated one-to-one to construct the extraction hierarchical optimization sub-strategy library; High-level extraction is performed on the M sets of partitioned historical extraction state vectors to determine M high-level extraction state vectors, including: Calculate the mean of the M sets of historical extraction state vectors to determine the M standard condition historical extraction state vectors; According to the preset extraction vector tolerance bandwidth, the M standard condition historical extraction state vectors are extracted and iterated in random directions in the M partitioned historical extraction state vector sets to determine the M iterated historical extraction state vectors. The update iteration direction is determined based on the M iterative history extraction state vectors and the M standard condition history extraction state vectors. High-level extraction is performed based on the update iteration direction to determine the M high-level extraction state vectors. Based on two dimensions—score magnitude and action variable clustering—and combining the M partitioned historical extracted state-action mapping score sets, the state variables in the M partitioned historical extracted state vector sets are filtered to determine M low-level action variables, including: According to the preset aggregation bandwidth, the density of action variables gathered around each action variable in the M sets of partitioned historical extracted state vectors is statistically analyzed to determine the M sets of partitioned action variable aggregation degrees. Based on two dimensions—score magnitude and action variable clustering—the state variables in the M partitioned historical extracted state-action mapping score sets and the M partitioned action variable clustering sets are filtered to determine the M low-level action variables.

2. The method for optimizing and controlling the recycling process of optoelectronic display materials as described in claim 1, characterized in that, Based on the M iterative historical extraction state vectors and the M standard condition historical extraction state vectors, the update iteration direction is determined. High-level extraction is then performed based on the update iteration direction to determine M high-level extraction state vectors, including: The similarity between each of the M iterative historical extraction state vectors and the historical extraction state vectors in the corresponding partitioned historical extraction state vector set is calculated using the cosine similarity calculation formula. The number of results greater than or equal to the preset similarity threshold is counted to obtain the update coefficients of the M iterative historical extraction state vectors. The similarity between each historical extraction state vector of the M standard conditions and the historical extraction state vector in the corresponding partitioned historical extraction state vector set is calculated using the cosine similarity calculation formula. The number of results greater than or equal to the preset similarity threshold is counted to obtain the update coefficients of the M historical extraction state vectors of the standard conditions. When the update coefficients of the M iterative historical extraction state vectors are less than the update coefficients of the M standard condition historical extraction state vectors, the direction from the M iterative historical extraction state vectors to the M standard condition historical extraction state vectors is taken as the update iteration direction. According to the update iteration direction, the M standard condition historical extraction state vectors continue to be iterated according to the preset extraction vector tolerance bandwidth until the preset number of iterations is met, and M high-level extraction state vectors are obtained. When the update coefficients of the M iterative historical extracted state vectors are greater than or equal to the update coefficients of the M standard case historical extracted state vectors, the direction from the M standard case historical extracted state vectors to the M iterative historical extracted state vectors is taken as the update iteration direction. According to the update iteration direction, the M iterative historical extracted state vectors are iterated again according to the preset extraction vector tolerance bandwidth until the preset number of iterations is met, and M high-level extracted state vectors are obtained.

3. The method for optimizing and controlling the recycling process of optoelectronic display materials as described in claim 1, characterized in that, Obtain the target process node and its state variables, and combine them with the hierarchical optimization strategy library to execute regeneration process optimization control, thereby determining the target regeneration process optimization control scheme, including: Based on the target process node and the state variables of the target process node, the hierarchical optimization strategy library is matched with strategies to determine the matching action variables; The matching action variables are randomly adjusted multiple times to determine the set of adjusted action variables; Obtain the update state-action associated reliability prediction branch, and use the update state-action associated reliability prediction branch to perform reliability prediction on the target process node state variables and adjustment action variable set respectively, and determine the adjustment reliability set; The adjustment action variable corresponding to the maximum value in the set of adjustment reliability is taken as the adjustment direction, and the adjustment action variables corresponding to the remaining adjustment reliability are adjusted according to the preset adjustment range to determine the set of follow adjustment action variables and the set of follow adjustment reliability. When there is an adjustment reliability in the set of follow-up adjustment reliability that is greater than the adjustment reliability corresponding to the adjustment direction, the adjustment action variable corresponding to the maximum value in the set of follow-up adjustment reliability is used as the updated adjustment direction, and the remaining set of follow-up adjustment action variables is adjusted until the preset number of adjustments is met, the target adjustment action variable is obtained, and the target adjustment action variable is used as the target regeneration process optimization control scheme.

4. The method for optimizing and controlling the recycling process of optoelectronic display materials as described in claim 3, characterized in that, When there is no adjustment reliability greater than the adjustment reliability corresponding to the adjustment direction in the set of follow-up adjustment reliability, the set of follow-up adjustment action variables is adjusted in the adjustment direction according to the preset adjustment range until the preset number of adjustments is met, and the target adjustment action variable is obtained. The target adjustment action variable is used as the target regeneration process optimization control scheme.

5. The method for optimizing and controlling the recycling process of optoelectronic display materials as described in claim 3, characterized in that, The updated state-action association reliability prediction branch is obtained, including: Obtain the initial state-action association reliability prediction branch; The node index improvement variables and action variables within the historical extraction state vector set, historical recrystallization state vector set, historical purification state vector set, and historical drying state vector set are used to update and adjust the parameters of the initial state-action correlation reliability prediction branch, thereby obtaining the updated state-action correlation reliability prediction branch after parameter update.

6. An optimization control system for the recycling process of optoelectronic display materials, characterized in that, The system is used to execute the optoelectronic display material regeneration process optimization and control method according to any one of claims 1-5, the system comprising: Historical data acquisition module: Acquires the set of key regeneration process nodes of the target optoelectronic display material, traverses the set of key regeneration process nodes to collect the state vectors of historical key regeneration process nodes, and obtains the set of historical extraction state vectors, the set of historical recrystallization state vectors, the set of historical purification state vectors and the set of historical drying state vectors. Each state vector includes state variables, action variables and node index improvement variables. Association mapping module: Iterates through the set of historical extraction state vectors, the set of historical recrystallization state vectors, the set of historical purification state vectors, and the set of historical drying state vectors, performs association mapping scoring on the state variables and action variables based on the node indicator improvement variables, and constructs a hierarchical optimization strategy library based on the scoring results; Optimization control module: acquires the target process node and the target process node state variables, executes regeneration process optimization control in conjunction with the hierarchical optimization strategy library, determines the target regeneration process optimization control scheme, and optimizes the target process node based on the target regeneration process optimization control scheme.

Images