Manufacturing method of packaging substrate of LED electronic display screen

Abstract

The invention discloses a manufacturing method of a packaging substrate of an LED electronic display screen, and relates to the technical field of LED circuit board processing. Through integration of a material demand plan MRP system and CNC cutting equipment and utilization of an intelligent typesetting algorithm based on a genetic algorithm, the utilization of substrate materials is maximized, and waste is reduced; an intelligent drilling adjustment system is constructed by adopting a depth Q learning algorithm, real-time adjustment and optimization of drilling parameters are realized, and the hole position precision and the hole shape consistency are ensured; in the intelligent plating process, electroplating parameters are monitored and adjusted in real time by utilizing a neural network model based on deep learning, the uniformity and adhesive force of an electroplated layer are improved, and finally, an automatic plate grinding machine is adopted for resin grinding in the press-fit brownification and resin grinding stages, so that the flatness of the surface of a substrate is ensured.

Description

Technical Field

[0001] The invention relates to the technical field of LED circuit board processing, in particular to a method for manufacturing a packaging substrate of an LED electronic display screen. Background Art

[0002] The packaging process of LED electronic display screens is to protect LED chips and circuit components, provide good heat dissipation performance, and facilitate installation and maintenance; common LED display screen packaging forms include SMD packaging, lamp bead packaging, packaging module, and COB packaging. In SMD packaging, the LED chip is directly mounted on the PCB, connected to the circuit by welding, and then encapsulated with transparent epoxy resin. It has the advantages of small size, light weight, and easy installation. In lamp bead packaging, the LED chip is installed in a small plastic or metal shell, usually with two pins. This packaging method is often used in outdoor LED displays. Due to its relatively strong structure, it can better withstand harsh environmental conditions. The packaging module is a larger and more complete packaging form, including LED chips, drive circuits, power supplies and other components to form a fully functional LED display module. The packaging module is usually installed on a bracket or frame to build the entire LED display. In COB packaging, multiple LED chips are directly mounted on the PCB and connected by conductive glue. COB packaging has the characteristics of high density and high brightness.

[0003] For example, the patent with publication number CN111263518B discloses a method for manufacturing a packaging substrate for an LED electronic display screen, including the following steps: cutting → drilling → VCP electroplating → pressing browning → solder mask plugging → baking and curing → resin grinding → outer layer dry film → etching → outer layer AOI inspection → solder mask → solder mask curing → sandblasting → electroplating nickel silver gold → CNC numerical control molding → finished product cleaning → final inspection → reliability → packaging. Through the processing method of the present invention, the drilling hole position accuracy of ±0.05mm can be guaranteed, and the mass-produced products pass the solderability, hot and cold shock, heat shock resistance, bonding gold wire, and gold ball thrust tests.

[0004] Although the above scheme has the above advantages, the traditional LED electronic display packaging substrate manufacturing method usually uses manual layout and manual adjustment of parameters, which is prone to uneven layout and serious material waste; and in the drilling and electroplating processes, there is a lack of real-time monitoring and adjustment, resulting in the inability to effectively guarantee the hole position accuracy and coating quality. In addition, the production process of the traditional method is relatively cumbersome, relying on experience and manual operation, and is prone to human errors, affecting production efficiency and product quality. Therefore, a method for manufacturing a LED electronic display packaging substrate is urgently needed to solve such problems. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention provides a method for manufacturing a packaging substrate of an LED electronic display screen, which solves the problems of manual typesetting and manual parameter adjustment in the prior art, which easily lead to uneven typesetting and serious waste of materials.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] The present invention provides a method for manufacturing a packaging substrate of an LED electronic display screen, comprising:

[0008] Step 1. Perform CNC cutting, integrate the material requirement planning MRP system with the CNC cutting equipment, and perform intelligent typesetting based on production orders and material inventory;

[0009] Step 2. Perform intelligent drilling, monitor the drilling process in real time, and use machine learning and image recognition technology to make real-time adjustments based on the monitoring results; determine the drilling parameters, including rotation speed, tool entry speed, and tool retraction speed, to optimize the hole position accuracy and hole shape consistency; detect and correct hole position deviations;

[0010] Step 3. Perform intelligent plating, use intelligent control algorithms to control the electroplating process, monitor and adjust the temperature, concentration and current density parameters of the plating solution in real time during electroplating; automatically adjust the parameters of the electroplating process according to the feedback data to adjust the uniformity and adhesion of the electroplating layer;

[0011] Step 4. Pressing and browning to adjust the surface flatness and corrosion resistance of the substrate;

[0012] Step 5. Resin grinding and finished product cleaning, use an automatic grinding machine to grind the resin to ensure the flatness of the substrate surface.

[0013] The present invention is further configured as follows: the intelligent typesetting method in step 1 is:

[0014] Obtain production order information O, including part quantity, size and demand date, and material inventory S, including available substrate quantity, size and material type, through the material requirement planning MRP system;

[0015] Preprocess and clean the data to remove invalid data and duplicates;

[0016] The present invention is further configured as follows: the intelligent typesetting method in step 1 further includes:

[0017] Construct an intelligent layout algorithm based on genetic algorithm, aiming to maximize the utilization of substrate materials and use binary coding to represent substrate layout;

[0018] Assume that the number of parts that each substrate j can accommodate is C j, the size of each part i is L i ×W i , the typesetting scheme X is represented by (x ij ), where x ij Indicates whether part i is placed on substrate j, and the value is 0 or 1;

[0019] The optimization goal is to maximize the utilization ratio U, which is the ratio of the total part area to the total substrate area;

[0020] The present invention is further configured as follows: in step 1, the optimization target maximizing utilization is specifically:

[0021]

[0022] The constraint is that each part can only be placed on one substrate:

[0023] The total area of ​​each base plate cannot exceed its capacity: Where U is the utilization rate, which represents the ratio of the total area of ​​all placed parts to the total area of ​​all substrates, and is used to measure the utilization efficiency of the substrate. M is the number of parts, which represents the number of parts that need to be processed in the production order. N is the number of substrates, which represents the number of substrates that can be used for typesetting. ij is a binary variable, indicating whether part i is placed on substrate j, L i and W i Respectively represent the length and width of part i, L j and W j is the length and width of substrate j, i is the index of the part, ranging from 1 to M, j is the index of the substrate, ranging from 1 to N;

[0024] The present invention is further configured as follows: in step 2, an intelligent drilling adjustment system is constructed to perform intelligent drilling, specifically:

[0025] Use high-speed cameras and sensors to monitor the drilling process in real time, collecting hole position deviation d, hole shape consistency c, and surface quality q;

[0026] Preprocess the collected data, including denoising, filtering and data cleaning;

[0027] Define the state space S, including monitoring parameters d, c, q, and extract additional features according to the actual application scenario, including ambient temperature T and drill wear degree W;

[0028] Define the action space A, including the adjustment of drilling processing parameters, rotation speed r, cutting speed v, retracting speed u, and the change of drill diameter D;

[0029] Define the reward function R to measure the quality and efficiency of each action, which is the weighted sum of processing quality;

[0030] Use the neural network model to model the state space S and action space A, and build an intelligent drilling adjustment system based on deep Q learning DQN;

[0031] Use historical data to train the model and use reinforcement learning algorithms to optimize model parameters;

[0032] Apply the trained model to the actual drilling process, monitor the processing status in real time, and adjust the drilling parameters according to the model output;

[0033] The present invention is further configured as follows: a method for constructing an intelligent drilling adjustment system based on deep Q learning DQN includes:

[0034] Build a deep Q learning DQN neural network model to approximate the action value function Q(s,a), where s is the state and a is the action;

[0035] The input layer of the neural network model takes the monitoring parameters as input, the hidden layer uses a multi-layer perceptron MLP structure, and the output layer outputs the Q value of each action;

[0036] Using the mean square error (MSE) loss function, the predicted Q value is compared with the target Q value, and the model parameters are updated through the back propagation algorithm.

[0037] Use the ε greedy strategy to balance exploration and exploitation, where ε is a positive number less than 1. At each time step, a random action is selected with probability ε, and the action with the highest Q value is selected with probability 1-ε.

[0038] The present invention is further configured as follows: a method for constructing an intelligent drilling adjustment system based on deep Q learning DQN also includes:

[0039] Calculate the Q value Q(s,a) of executing action a under the current state s, and get the predicted value through the neural network model;

[0040] Calculate the target Q value Q target (s,a) is used to update the neural network parameters according to the update rule of the Bellman equation:

[0042] Where r is the reward, γ is the discount factor, s' is the next state, and a' is the action under s';

[0043] The present invention is further configured as follows: a method for constructing an intelligent drilling adjustment system based on deep Q learning DQN also includes:

[0044] By interacting with the environment, experience data e is collected, including state, action, reward, and next state;

[0045] The experience data is stored in the experience replay buffer and randomly sampled from it to reduce the correlation between data;

[0046] Use stochastic gradient descent SGD to update the parameters of the neural network model through the back-propagation algorithm;

[0047] The present invention is further configured as follows: the electroplating process control method is:

[0048] Conduct electroplating parameter monitoring and data collection, monitoring the temperature T, concentration C and current density I during the electroplating process;

[0049] Construct a neural network model based on deep learning to model and predict the dynamic changes in the electroplating process. The model uses a convolutional neural network (CNN) combined with a long short-term memory (LSTM) network architecture.

[0050] The input of the neural network model is the monitoring data X={T,C,I} during the electroplating process, and the output of the neural network model is the dynamically adjusted electroplating parameters;

[0051] The neural network model was trained and optimized using the dominant action-critic algorithm A2C, and the electroplating parameters were adjusted by maximizing the reward function using a reinforcement learning framework;

[0052] The present invention is further configured as follows: the electroplating process control method further includes:

[0053] The reward function R(A) is designed to measure the quality and efficiency of electroplating parameter adjustment: Where: w U 、w F , are the weight coefficients of uniformity, adhesion and processing time, respectively. U represents the uniformity of the electroplating layer, F represents the adhesion of the electroplating layer, and T P Indicates the processing time of electroplating.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention introduces the integration of the material requirement planning system and the CNC cutting equipment and the intelligent typesetting algorithm based on the genetic algorithm to achieve automatic typesetting and maximize material utilization, greatly reducing material waste and improving production efficiency;

[0056] The present invention adopts an intelligent adjustment system based on deep learning during the drilling and electroplating process to achieve real-time monitoring and automatic adjustment, effectively improving the hole position accuracy, hole shape consistency, and uniformity and adhesion of the electroplating layer;

[0057] The invention solves the problems existing in the prior art of manual typesetting and manual parameter adjustment, which easily lead to uneven typesetting and serious waste of materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 The present invention is a flow chart of a method for manufacturing a packaging substrate of an LED electronic display screen. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0060] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0061] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0062] Example

[0063] See also Figure 1 The present invention provides a method for manufacturing a packaging substrate of an LED electronic display screen, comprising:

[0064] Step 1. Perform CNC cutting, integrate the material requirement planning MRP system with the CNC cutting equipment, and perform intelligent layout according to production orders and material inventory to maximize the use of substrate materials and reduce waste;

[0065] The intelligent typesetting method in step 1 is: obtaining production order information O, including part quantity, size and demand date, and material inventory S, including available substrate quantity, size and material type, through the material requirement planning MRP system;

[0066] Preprocess and clean the data to remove invalid data and duplicates;

[0067] Construct an intelligent layout algorithm based on genetic algorithm, aiming to maximize the utilization of substrate materials and use binary coding to represent substrate layout;

[0068] Assume that the number of parts that each substrate j can accommodate is C j , the size of each part i is L i ×W i , the typesetting scheme X is represented by (x ij ), where x ij Indicates whether part i is placed on substrate j, and the value is 0 or 1;

[0069] The optimization goal is to maximize the utilization ratio U, which is the ratio of the total part area to the total substrate area;

[0070] The optimization goal of maximizing utilization is as follows:

[0071] The constraint is that each part can only be placed on one substrate:

[0072] The total area of ​​each base plate cannot exceed its capacity: Where U is the utilization rate, which represents the ratio of the total area of ​​all placed parts to the total area of ​​all substrates, and is used to measure the utilization efficiency of the substrate. M is the number of parts, which represents the number of parts that need to be processed in the production order. N is the number of substrates, which represents the number of substrates that can be used for typesetting. ij is a binary variable, indicating whether part i is placed on substrate j, L i and W i Respectively represent the length and width of part i, L j and W j is the length and width of substrate j, i is the index of the part, ranging from 1 to M, j is the index of the substrate, ranging from 1 to N;

[0073] The objective function aims to maximize the utilization of the substrate. The constraints ensure that each part can only be placed on one substrate and the total area of ​​each substrate cannot exceed its capacity. Finally, a genetic algorithm is used for intelligent layout to iteratively find the optimal solution, thereby maximizing the utilization of substrate materials and reducing waste.

[0074] Step 2. Perform intelligent drilling, monitor the drilling process in real time, and use machine learning and image recognition technology to make real-time adjustments based on the monitoring results; determine the drilling parameters, including rotation speed, cutting speed, and retracting speed, to optimize the hole position accuracy and hole shape consistency; detect and correct the hole position deviation, and combine machine learning and real-time control algorithms to achieve intelligent optimization and real-time adjustment of the drilling process;

[0075] In step 2, an intelligent drilling adjustment system is constructed to perform intelligent drilling, specifically:

[0076] Use high-speed cameras and sensors to monitor the drilling process in real time, collecting hole position deviation d, hole shape consistency c, and surface quality q;

[0077] Preprocess the collected data, including denoising, filtering and data cleaning;

[0078] Define the state space S, including monitoring parameters d, c, q, and extract additional features according to the actual application scenario, including ambient temperature T and drill wear degree W;

[0079] Define the action space A, including the adjustment of drilling processing parameters, rotation speed r, cutting speed v, retracting speed u, and the change of drill diameter D;

[0080] Define the reward function R to measure the quality and efficiency of each action, which is the weighted sum of processing quality;

[0081] Use the neural network model to model the state space S and action space A, and build an intelligent drilling adjustment system based on deep Q learning DQN;

[0082] Use historical data to train the model and use reinforcement learning algorithms to optimize model parameters;

[0083] Apply the trained model to the actual drilling process, monitor the processing status in real time, and adjust the drilling parameters according to the model output to ensure the hole position accuracy and hole shape consistency;

[0084] Build a deep Q learning DQN neural network model to approximate the action value function Q(s,a), where s is the state and a is the action;

[0085] The input layer of the neural network model takes the monitoring parameters as input, the hidden layer uses a multi-layer perceptron MLP structure, and the output layer outputs the Q value of each action;

[0086] Using the mean square error (MSE) loss function, the predicted Q value is compared with the target Q value, and the model parameters are updated through the back propagation algorithm.

[0087] Use the ε greedy strategy to balance exploration and exploitation, where ε is a positive number less than 1. At each time step, a random action is selected with probability ε, and the action with the highest Q value is selected with probability 1-ε.

[0088] Calculate the Q value Q(s,a) of executing action a under the current state s, and get the predicted value through the neural network model;

[0089] Calculate the target Q value Q target (s,a) is used to update the neural network parameters according to the update rule of the Bellman equation: Where r is the reward, γ is the discount factor, s' is the next state, and a' is the action under s';

[0090] The methods of building an intelligent drilling adjustment system based on deep Q learning DQN also include:

[0091] By interacting with the environment, experience data e is collected, including state, action, reward, and next state;

[0092] The experience data is stored in the experience replay buffer and randomly sampled from it to reduce the correlation between data;

[0093] Use stochastic gradient descent SGD to update the parameters of the neural network model through the back-propagation algorithm;

[0094] This enables the design and training of an intelligent drilling adjustment system based on deep Q-learning to optimize the parameter settings of the drilling process;

[0095] Step 3. Perform intelligent plating, use intelligent control algorithms to control the electroplating process, monitor and adjust the temperature, concentration and current density parameters of the plating solution in real time during electroplating; automatically adjust the parameters of the electroplating process according to the feedback data to adjust the uniformity and adhesion of the electroplating layer;

[0096] The electroplating process control method is:

[0097] Conduct electroplating parameter monitoring and data collection, monitoring the temperature T, concentration C and current density I during the electroplating process;

[0098] Construct a neural network model based on deep learning to model and predict the dynamic changes in the electroplating process. The model uses a convolutional neural network (CNN) combined with a long short-term memory (LSTM) network architecture.

[0099] The input of the neural network model is the monitoring data X={T,C,I} during the electroplating process, and the output of the neural network model is the dynamically adjusted electroplating parameters;

[0100] The neural network model is trained and optimized using the dominant action-criteria algorithm A2C, and the electroplating parameters are adjusted by maximizing the reward function using the reinforcement learning framework to improve the uniformity and adhesion of the electroplated layer.

[0101] The reward function R(A) is designed to measure the quality and efficiency of electroplating parameter adjustment:, where: w U 、w F , are the weight coefficients of uniformity, adhesion and processing time, respectively. U represents the uniformity of the electroplating layer, F represents the adhesion of the electroplating layer, and T P Indicates the processing time of electroplating;

[0102] Step 4. Pressing and browning to adjust the surface flatness and corrosion resistance of the substrate;

[0103] Step 5. Resin grinding and finished product cleaning, use an automatic grinding machine to grind the resin to ensure the flatness of the substrate surface.

[0104] The present invention, through the integration of the material requirement planning MRP system and the CNC cutting equipment, utilizes the intelligent typesetting algorithm based on the genetic algorithm to maximize the utilization of substrate materials and reduce waste. Secondly, in the intelligent drilling stage, an intelligent drilling adjustment system constructed by a deep Q learning algorithm is adopted to realize real-time adjustment and optimization of drilling parameters to ensure hole position accuracy and hole shape consistency. In the intelligent plating process, a neural network model based on deep learning is used to monitor and adjust the electroplating parameters in real time to improve the uniformity and adhesion of the electroplating layer. Finally, in the pressing browning and resin grinding stages, an automatic plate grinding machine is used for resin grinding to ensure the flatness of the substrate surface.

[0105] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for manufacturing a packaging substrate for an LED electronic display screen, characterized in that: include: Step 1. Perform CNC cutting, integrate the material requirement planning MRP system with the CNC cutting equipment, and perform intelligent typesetting based on production orders and material inventory; Step 2. Perform intelligent drilling, monitor the drilling process in real time, and use machine learning and image recognition technology to make real-time adjustments based on the monitoring results; determine the drilling parameters, including rotation speed, tool entry speed, and tool retraction speed, to optimize the hole position accuracy and hole shape consistency; detect and correct hole position deviations; Step 3. Perform intelligent plating, use intelligent control algorithms to control the electroplating process, monitor and adjust the temperature, concentration and current density parameters of the plating solution in real time during electroplating; automatically adjust the parameters of the electroplating process according to the feedback data to adjust the uniformity and adhesion of the electroplating layer; Step 4. Pressing and browning to adjust the surface flatness and corrosion resistance of the substrate; Step 5. Resin grinding and finished product cleaning, use an automatic grinding machine to grind the resin to ensure the flatness of the substrate surface.

2. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 1, characterized in that: The smart layout method in step 1 is: Obtain production order information O, including part quantity, size and demand date, and material inventory S, including available substrate quantity, size and material type, through the material requirement planning MRP system; Preprocess and clean the data to remove invalid data and duplicates.

3. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 2, characterized in that: The smart typesetting methods in step 1 also include: Construct an intelligent layout algorithm based on genetic algorithm, aiming to maximize the utilization of substrate materials and use binary coding to represent substrate layout; Assume that the number of parts that each substrate j can accommodate is C j , the size of each part i is L i ×W i , the typesetting scheme X is represented by (x ij ), where x ij Indicates whether part i is placed on substrate j, and the value is 0 or 1; The optimization goal is to maximize the utilization ratio U, which is the ratio of the total part area to the total substrate area.

4. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 3, characterized in that: In step 1, the optimization objective of maximizing utilization is as follows: The constraint is that each part can only be placed on one substrate: The total area of ​​each base plate cannot exceed its capacity: Where U is the utilization rate, which represents the ratio of the total area of ​​all placed parts to the total area of ​​all substrates, and is used to measure the utilization efficiency of the substrate. M is the number of parts, which represents the number of parts that need to be processed in the production order. N is the number of substrates, which represents the number of substrates that can be used for typesetting. ij is a binary variable, indicating whether part i is placed on substrate j, L i and W i Respectively represent the length and width of part i, L j and W j are the length and width of substrate j, i is the index of the part, ranging from 1 to M, and j is the index of the substrate, ranging from 1 to N.

5. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 4, characterized in that: In step 2, an intelligent drilling adjustment system is constructed to perform intelligent drilling, specifically: Use high-speed cameras and sensors to monitor the drilling process in real time, collecting hole position deviation d, hole shape consistency c, and surface quality q; Preprocess the collected data, including denoising, filtering and data cleaning; Define the state space S, including monitoring parameters d, c, q, and extract additional features according to the actual application scenario, including ambient temperature T and drill wear degree W; Define the action space A, including the adjustment of drilling processing parameters, rotation speed r, cutting speed v, retracting speed u, and the change of drill diameter D; Define the reward function R to measure the quality and efficiency of each action, which is the weighted sum of processing quality; Use the neural network model to model the state space S and action space A, and build an intelligent drilling adjustment system based on deep Q learning DQN; Use historical data to train the model and use reinforcement learning algorithms to optimize model parameters; The trained model is applied to the actual drilling process, the processing status is monitored in real time, and the drilling parameters are adjusted according to the model output.

6. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 5, characterized in that: The methods for building an intelligent drilling adjustment system based on deep Q learning DQN include: Build a deep Q learning DQN neural network model to approximate the action value function Q(s,a), where s is the state and a is the action; The input layer of the neural network model takes the monitoring parameters as input, the hidden layer uses a multi-layer perceptron MLP structure, and the output layer outputs the Q value of each action; Using the mean square error (MSE) loss function, the predicted Q value is compared with the target Q value, and the model parameters are updated through the back propagation algorithm. Use the ε-greedy strategy to balance exploration and exploitation, where ε is a positive number less than 1. At each time step, a random action is selected with probability ε, and the action with the highest Q value is selected with probability 1-ε.

7. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 6, characterized in that: The methods of building an intelligent drilling adjustment system based on deep Q learning DQN also include: Calculate the Q value Q(s,a) of executing action a under the current state s, and get the predicted value through the neural network model; Calculate the target Q value Q target (s,a) is used to update the neural network parameters according to the update rule of the Bellman equation: Q target (s,a)=r+γmax a' Q(s′,a′), where r is the reward, γ is the discount factor, s′ is the next state, and a′ is the action under s′.

8. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 7, characterized in that: The methods of building an intelligent drilling adjustment system based on deep Q learning DQN also include: By interacting with the environment, experience data e is collected, including state, action, reward, and next state; The experience data is stored in the experience replay buffer and randomly sampled from it to reduce the correlation between data; Stochastic gradient descent (SGD) is used to update the parameters of the neural network model through the back-propagation algorithm.

9. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 8, characterized in that: The electroplating process control method is: Conduct electroplating parameter monitoring and data collection, monitoring the temperature T, concentration C and current density I during the electroplating process; Construct a neural network model based on deep learning to model and predict the dynamic changes in the electroplating process. The model uses a convolutional neural network (CNN) combined with a long short-term memory (LSTM) network architecture. The input of the neural network model is the monitoring data X={T,C,I} during the electroplating process, and the output of the neural network model is the dynamically adjusted electroplating parameters; The neural network model was trained and optimized using the advantage action-critic algorithm A2C, and the electroplating parameters were adjusted by maximizing the reward function using a reinforcement learning framework.

10. The method for manufacturing a packaging substrate of an LED electronic display screen according to claim 9, characterized in that: Electroplating process control methods also include: The reward function R(A) is designed to measure the quality and efficiency of electroplating parameter adjustment: Where: w U 、w F , are the weight coefficients of uniformity, adhesion and processing time, respectively. U represents the uniformity of the electroplating layer, F represents the adhesion of the electroplating layer, and T P Indicates the processing time of electroplating.

Images