Production process optimization control method and system for shading adhesive tape

Abstract

The invention provides a shading tape production process optimization control method and system, and belongs to the technical field of industrial control, and the method comprises the steps: S100, obtaining a production process associated data set, carrying out the key information extraction of the production process associated data set, and building a production process associated map; step S200, acquiring a real-time production process control set, constructing a production process intelligent agent cluster based on the real-time production process control set and the production process association map, and generating a production process optimization control strategy through an intelligent agent cluster virtual optimization strategy; s300, performing digital twinborn verification on the production process optimization control strategy to generate a production process final control strategy, and performing real-time regulation and control based on the production process final control strategy; and S400, obtaining the regulated and controlled associated data, and updating the production process associated map based on the regulated and controlled associated data. The method provides a control method for global and multi-directional optimization of the shading tape production process.

Description

Technical Field

[0001] The present invention relates to the field of industrial control technology, and in particular to a production process optimization control method and system for light-shielding tape. Background Art

[0002] In modern industrial production and daily life, blackout tape, a composite material that combines shading, bonding, and insulation functions, is finding increasingly widespread application. In electronic equipment, it's used to seal the borders of liquid crystal displays and shield backlight modules, ensuring the screen's display is uninterrupted by stray light. In automotive manufacturing, it's often used to secure headlight assemblies to ensure accurate and safe driving lighting. In the architectural decoration industry, it can be used in conjunction with glass curtain walls to achieve partial shading and adjust indoor lighting. As downstream industries continue to demand more precision and stability from products, the quality standards for blackout tape are becoming increasingly stringent. However, the production process control for blackout tape remains limited.

[0003] On the one hand, single-process parameter tuning, such as PID control of coating pressure, often ignores the coupling effect of multiple processes such as coating-curing-slitting, resulting in local optimization leading to global deterioration. For example, high coating pressure improves thickness uniformity but increases oven energy consumption.

[0004] On the other hand, conventional digital twin verification relies on historical operating data, has low coverage of extreme disturbances, and often cannot warn of compound risks, resulting in a lack of risk assurance in the generated optimized control instructions.

[0005] Therefore, it is urgent to build a new generation of global optimization framework for production processes that integrates swarm intelligent game and adversarial twin verification to realize the self-perception and self-decision-making global optimization control method of the entire link of the light-shielding tape production process. Summary of the Invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for optimizing and controlling the production process of a light-shielding tape, the method comprising: S100, obtaining a production process related data set, extracting key information from the production process related data set, and establishing a production process related map; S200, obtaining a real-time production process control set, building a production process agent cluster based on the real-time production process control set and the production process association map, and generating a production process optimization control strategy through an agent cluster virtual optimization strategy; S300, performing digital twin verification on the production process optimization control strategy, generating a final production process control strategy, and performing real-time regulation based on the final production process control strategy; S400: Acquire post-control correlation data, and update the production process correlation map based on the post-control correlation data.

[0007] As a further solution of the present invention, obtaining a production process related data set, extracting key information from the production process related data set, and establishing a production process related map include: acquiring a production process related data set based on the production monitoring equipment group, and performing data processing operations on the production process related data set; Extracting key indicators of production processes from a production process-related data set through a neural network, and establishing a key production process matrix based on the key indicators of production processes; Calculating the production linear weight, production nonlinear weight, and production time lag weight based on the production process key matrix, importing the production process key matrix into the PC algorithm, and generating a basic production process association map based on the PC algorithm; The production process association map is generated by the basic production process association map, the production linear weight, the production nonlinear weight and the production lag weight.

[0008] As a further solution of the present invention, a real-time production process control set is obtained, a production process agent cluster is constructed based on the real-time production process control set and the production process association map, and a production process optimization control strategy is generated through the agent cluster virtual optimization strategy, including: Initialize the agent environment based on the real-time production process control set and the production process association map to generate a production process agent cluster; Acquire process agent observation data based on the production process agent cluster, process the process agent observation data based on a triple specificity algorithm, and acquire an original production process control strategy; Performing secondary optimization on the original production process control strategy to obtain a production process optimization strategy; Acquire a real-time production process association set, define a production utility objective function based on the real-time production process association set and the production process optimization strategy, and acquire a multidimensional utility vector based on the production utility objective function; A collaborative control algorithm is executed based on the multidimensional utility vector to obtain a multidimensional collaborative utility vector, and the production utility objective function is adjusted based on the multidimensional collaborative utility vector to obtain a production process optimization control strategy.

[0009] As a further solution of the present invention, obtaining a production process optimization strategy includes: Based on the real-time production process control set and the production process association map, a first production process agent, a second production process agent, and a third production process agent are defined and combined into the production process agent cluster; based on the production process agent cluster, observation state data of the first agent, observation state data of the second agent, and observation data of the third agent are obtained; Import the observation state data of the first agent into the PSO algorithm, import the observation state data of the second agent into the DDPG algorithm, and import the observation data of the third agent into the Q-learning algorithm to obtain the original production process control strategy; The original production process control strategy is optimized secondary based on the Nash Q-learning algorithm to obtain a pending production process optimization strategy, obtain physical constraint data, and verify the pending production process optimization strategy based on the physical constraint data. After the verification is completed, the production process optimization strategy is obtained.

[0010] As a further solution of the present invention, the undetermined production process optimization strategy is verified based on the physical constraint data. After the verification is completed, the original production process control strategy is obtained, including: Obtaining physical constraint conditions based on physical constraint data; If the undetermined production process optimization strategy satisfies the physical constraint condition, redefining the undetermined production process optimization strategy as the production process control original strategy; If the pending production process optimization strategy does not satisfy the physical constraint conditions, the pending production process optimization strategy is modified until the physical constraint conditions are satisfied, and the original production process control strategy is obtained.

[0011] As a further solution of the present invention, obtaining a multidimensional utility vector based on the production utility objective function includes: Acquiring first utility data, second utility data, and third utility data based on the production utility objective function, and generating a multidimensional utility vector based on the first utility data, the second utility data, and the third utility data; Importing the multidimensional utility vector into a GA algorithm, generating an initial utility population based on the GA algorithm, performing mutation and crossover operations on the initial utility population to obtain a secondary utility population, screening the secondary utility population based on a tournament strategy, and after the screening is completed, outputting a pending multidimensional utility vector based on the GA algorithm; N groups of data are extracted from the quadratic utility population as a strategy candidate set. The strategy candidate set and the undetermined multidimensional utility vector are introduced into the NSGA-II algorithm. The undetermined multidimensional utility vector is optimized based on the NSGA-II algorithm to obtain a multidimensional utility vector.

[0012] As a further solution of the present invention, the production process optimization control strategy is digitally verified to generate a final production process control strategy, including: Constructing a virtual production process condition network, importing the production process optimization control strategy into the virtual production process condition network, and generating a virtual condition data set based on the virtual production process condition network; Constructing a production process condition prediction network, importing the virtual condition data set and the production process optimization control strategy into the production process condition prediction network, and obtaining a strategy-quality prediction mapping table based on the production process condition prediction network; Obtain historical working condition defect probabilities, construct a production process working condition entropy value calculation model based on the historical working condition defect probabilities and a strategy-quality prediction mapping table, and obtain a strategy risk entropy data set based on the production process working condition entropy value calculation model; A control strategy decision surface is established, and the strategy risk entropy data set, strategy-quality prediction mapping table and production process optimization control strategy are imported into the control strategy decision surface, and the final control strategy of the production process is generated based on the control strategy decision surface.

[0013] As a further embodiment of the present invention, the method further comprises: The production process condition network is established based on the GAN network and includes a generator, a discriminator and a loss function. The generator is constructed by the ResNet-18 network, and the discriminator is constructed by the GAN network. The production process condition prediction network is established based on the LSTM network and includes dual channels, one set of dual channels is a process channel, and the other set of dual channels is a quality channel; The control strategy decision surface is established based on the B-spline surface.

[0014] In another aspect, an embodiment of the present invention further provides a production process optimization control system for light-shielding tape, the system comprising: The atlas module is used to obtain a production process related data set, extract key information in the production process related data set, and construct a production process related atlas.

[0015] A primary strategy module, which is used to obtain a real-time production process control set, construct a production process agent cluster of the real-time production process control set and a production process association map, and generate a production process optimization control strategy through a virtual optimization strategy of the agent cluster; A secondary strategy module, which is used to perform digital twin verification and generate a final control strategy for the production process; A control module, which performs real-time control based on the final control strategy of the production process; The reconstruction module is used to obtain the associated data after regulation and update the production process association map.

[0016] Based on the above aspects, the embodiment of the present application first obtains a production-related data set and extracts key information from the production process-related data set to construct a production process-related map, providing low-dimensional and interpretable physical constraint input for the processing of subsequent steps. Secondly, by obtaining a real-time production process control set and constructing a production process process intelligent agent cluster of the real-time production process control set and the production process-related map, the different types of observation state data in the production process process intelligent agent cluster can be targetedly processed through a combined targeted algorithm of the PSO algorithm, the DDPG algorithm and the Q-learing algorithm, so as to provide an initial strategy value. Afterwards, by optimizing the production process The control strategy is verified by digital twins to generate the final control strategy for the production process, thereby eliminating low-quality strategies in the initial strategy and obtaining the optimal control strategy based on this. Finally, the production process association map can be updated based on the associated data after regulation to achieve global control of the light-shielding tape process and avoid the situation where local deterioration is caused by the traditional single-process parameter tuning. In this way, it adapts to the coupling effect. Through digital twin verification, risky strategies that may appear in the initial strategy can be comprehensively eliminated to warn of complex risks and ensure that the optimal control strategy finally output has global control of the process and a control strategy with risk assurance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention provides a flowchart of a method for optimizing the production process of a light-shielding tape.

[0018] Figure 2 It is a schematic diagram of the execution flow of step S200 in a method for optimizing and controlling the production process of a light-shielding tape provided in an embodiment of the present invention.

[0019] Figure 3 Schematic diagram of a production process optimization control system for a light-shielding tape provided by an embodiment of the present invention.

[0020] Figure 4 It is a schematic diagram of an execution flow of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The drawings in the embodiments clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0023] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 1 is a schematic diagram of an execution flow of a method for optimizing and controlling a production process of a light-shielding tape provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of the execution flow of step S200 in a method for optimizing the control of a production process of a light-shielding tape provided by an embodiment of the present invention. The method for optimizing the control of a production process of a light-shielding tape is introduced in detail below.

[0024] Specifically, a method for optimizing and controlling the production process of a light-shielding tape includes: Step S100: Acquire a production process related data set, extract key information from the production process related data set, and establish a production process related map.

[0025] Furthermore, the step S100 includes the following specific steps: Step S1000: obtaining a production process related data set based on a production monitoring equipment group, and performing data processing operations on the production process related data set.

[0026] Step S1001: extracting key indicators of production processes from a production process related data set through a neural network, and establishing a production process key matrix based on the key indicators of production processes.

[0027] Step S1002, calculating the production linear weight, production nonlinear weight and production lag weight based on the production process key matrix, and importing the production process key matrix into the PC algorithm, and generating a basic production process association map based on the PC algorithm.

[0028] Step S1003 , generating the production process association map through the basic production process association map, the production linear weight, the production nonlinear weight, and the production lag weight.

[0029] It can be understood that the process flow of the light-shielding tape may include process flows such as raw material preparation and processing, coating processing, aging processing, and die-cutting processing. The above four basic process flows are explained here. For raw material preparation and processing, the substrate is selected according to the product requirements and the black adhesive is prepared or selected according to the final use of the light-shielding tape. For coating processing, the prepared black adhesive is evenly coated on the surface of the substrate to form a light-shielding function. During the coating process, the coating thickness needs to be controlled, which directly affects the light-shielding effect. It is then dried and cured. During the drying and curing process, it also needs to be controlled to prevent bubbling and cracking. For aging processing, the composite material is stored to ensure that the adhesive completes the final cross-linking reaction. For die-cutting processing, the aging material is cut on a slitting machine according to customer needs. This method is established based on the basic coating, curing, and cutting production process. In actual use, it can be adjusted and expanded based on the actual production process of the light-shielding tape to match different implementation methods.

[0030] It can be understood that the multi-source data in the production process of the light-shielding tape is monitored by a production monitoring equipment group. The production monitoring equipment may include industrial high-speed cameras, vibration sensors, temperature and humidity sensors, infrared thermal imagers, etc., and the multi-source data collected by the monitoring may include surface image data collected by industrial high-speed cameras, vibration spectrum data of motor rotation collected by vibration sensors, temperature gradient data of ovens collected by infrared thermal imagers, etc. The multi-source data collected by the production monitoring equipment are organized into a production process-related data set, and then data processing operations can be performed on the production process-related data set.

[0031] Furthermore, the production process-related data sets are aligned in time and space to provide a unified benchmark data foundation for subsequent processes, solve the time misalignment and spatial offset problems of multi-source production process-related data sets, and can be processed through a clock synchronization network.

[0032] In this embodiment, the data collected and monitored by production monitoring equipment such as industrial high-speed cameras and vibration sensors can be synchronized to within a satisfactory error range through a clock synchronization network such as the IEEE 1588 PTP, eliminating the timing confusion caused by protocol errors in traditional systems. For the interface protocols used by different production monitoring equipment, a dual-regulation mechanism of hardware trigger alignment and software timestamp injection can be adopted. For example, a high-speed camera uses a 2xxFPS image stream, and a vibration sensor uses a 5xkS / s spectrum, so that the time dimension strictly matches the displacement coordinates of the conveyor belt used in production. At the same time, a dynamic spatial mapping model is established to calculate the spatial offset of the data of each equipment station in real time based on the speed of the conveyor belt used in production. For example, the thickness measurement point at the coater outlet and the temperature measurement point at the oven inlet are overlapped in the coordinate system of the virtual production line. Ultimately, the timestamps of the multi-source production process-related data sets are consistent, providing a temporally and spatially consistent observation benchmark for subsequent production process optimization.

[0033] Furthermore, after the production process-related data sets are aligned in time and space, data cleaning is required, such as eliminating data on temperature jumps and vibration saturation. For example, if the temperature rise of adjacent sampling points is greater than 5 degrees Celsius, they will be marked as abnormal and eliminated. Such situations may be caused by poor contact of thermocouples. Abnormal data from the production monitoring equipment group is eliminated to eliminate distorted data from the source. After elimination, related data will be missing. At this time, it needs to be filled. For this, linear interpolation filling, ARIMA time series prediction model reconstruction, LSTM time series prediction model reconstruction, etc. can be used to fill the missing data to ensure data integrity.

[0034] In some possible embodiments, the data may also be normalized, such as by using a Softmax function. The Softmax function converts each element in the channel attention weight vector into a probability value. These probability values ​​represent the relative importance of each channel in the fusion process, thereby normalizing the data to a uniform interval and simplifying subsequent processing.

[0035] Specifically, neural networks can be used to refine multi-source, high-dimensional production process-related data sets and quantify them into quantifiable and interpretable process state features. Through neural network-driven feature engineering, feature extraction is performed on the processed production process-related data. The extractable features may include, for example, calculating the standard deviation of thickness measurement points to quantify coating uniformity, extracting the temperature gradient of the oven to characterize the balance of heat distribution, or extracting the vibration energy integral of the xxx-xxxHz frequency band in the vibration spectrum, etc. Through the above-mentioned feature extraction, multi-source, high-dimensional production process-related data sets can be converted into multi-source, low-dimensional production process key indicators. Then, the production process key indicators are combined to establish a production process key matrix.

[0036] It can be understood that by extracting the key indicators of the production process from the production process association data set, the multi-source high-dimensional data can be reduced in dimension while retaining the physical meaning of the process, and the original data can be distilled into key data features, providing low-dimensional and interpretable input for the subsequent establishment of the production process association map.

[0037] Specifically, before executing step S1002, corresponding historical production defect data is imported so that a map of the quantitative causal relationship between process-related parameters and quality defects can be established to achieve an interpretable data foundation.

[0038] Furthermore, the production linear weight, production nonlinear weight and production time lag weight are used to characterize the causal chain type of the subsequently established map. For example, the correlation between the temperature rise rate and the excessive light transmittance in the historical production defect data is a production nonlinear relationship, and its production linear weight is obtained. For example, the coating gap and the edge warping probability are a production linear relationship, and its production nonlinear weight is obtained. For example, the bubble generation caused by increased vibration has a delayed effect, and there is a delay between the increased vibration and the bubble generation, which is a production time lag relationship, and its production time lag weight is obtained.

[0039] It can be understood that the PC algorithm used in this embodiment is a causal discovery method based on conditional independence testing. It is used to identify the causal network structure between variables from observational data, which can overcome the limitation of traditional statistical methods that can only determine correlation. In specific use, the production process key matrix and historical production defect data are imported into the PC algorithm. First, a completely undirected graph is constructed, that is, all variables are connected in pairs. Then, the minimum independent set is found by stepwise deletion. For a pair of variables (X, Y), if there exists a condition set Z such that X⊥Y|Z (X and Y are independent when Z is given), the XY edge is deleted. Secondly, the V structure rule is used to determine partial causal orientation. That is, when X→Z←Y and X and Y have no direct connection, if X⊥Y but X does not⊥Y|Z, a causal collision is formed. Finally, the global causal direction is determined based on the directional propagation rule, that is, new V structures and loops are avoided. The algorithm outputs a basic production process association map by quantifying conditional independence.

[0040] It can be understood that through step S1000, the multi-source data of the production process-related data set can be aligned in time and space to solve the time sequence confusion caused by protocol errors and spatial errors in traditional systems. Secondly, abnormal data points in multi-source data can be eliminated through data cleaning, and the eliminated data points can be reasonably filled using the filling method. Finally, the causal chain is obtained based on the PC algorithm, and different linear weights, nonlinear weights and time lag weights are quantified, and a production process association map is constructed to provide a consistent and explainable data foundation with a time and space benchmark for subsequent intelligent optimization.

[0041] Step S200: obtain a real-time production process control set, construct a production process process intelligent agent cluster based on the real-time production process control set and the production process association map, and generate a production process optimization control strategy through the intelligent agent cluster virtual optimization strategy.

[0042] Furthermore, step S200 includes the following specific steps: Step S2000: Initialize the intelligent agent environment based on the real-time production process control set and the production process association map to generate a production process process intelligent agent cluster.

[0043] Step S2001: Acquire process agent observation data based on the production process procedure agent cluster, process the process agent observation data based on a triple specificity algorithm, and acquire the original production process control strategy.

[0044] Step S2002: perform secondary optimization on the original production process control strategy to obtain a production process optimization strategy.

[0045] Furthermore, step S2000 to step S2002 specifically include the following steps: Step S2000-1, based on the real-time production process control set and the production process association map, define the first intelligent agent of the production process, the second intelligent agent of the production process and the third intelligent agent of the production process, and combine them into the production process agent cluster, and obtain the first intelligent agent observation status data, the second intelligent agent observation status data, and the third intelligent agent observation data based on the production process agent cluster.

[0046] Specifically, the real-time production process control set may include relevant data during the process, such as real-time thickness data, real-time temperature data, real-time tension data, etc. The first intelligent agent of the production process can be represented as a coating intelligent agent, the second intelligent agent of the production process can be represented as a solidification intelligent agent, and the third intelligent agent of the production process can be represented as a slitting intelligent agent. By analyzing the production process association map, the abstract process map knowledge is converted into a mathematical boundary that can be understood by the machine. Then, the real-time production process control set is integrated to dynamically generate observation status data to form a dual cognitive framework of physical rules and real-time status, namely, coating intelligent agent observation status data, solidification intelligent agent observation status data, and slitting intelligent agent status data, to provide cognitive data for subsequent algorithm processing. In this embodiment, coating, solidification, and slitting are taken as examples. In the specific use process, they can be redefined, and other intelligent agents can be added to build a production process intelligent agent cluster.

[0047] In step S2000-2, the observation state data of the first agent is imported into the PSO algorithm, the observation state data of the second agent is imported into the DDPG algorithm, and the observation data of the third agent is imported into the Q-learning algorithm to obtain the original production process control strategy.

[0048] Specifically, for the high-dimensional continuous parameter space of the coating process, the PSO algorithm (particle swarm algorithm) is adopted. Through the multiple combinations formed by N particles in a short time, the optimal solution is output with thickness uniformity and warping risk as the fitness function. For the timing control of the curing temperature field, the DDPG algorithm (deep deterministic policy gradient algorithm) is adopted. The fan frequency conversion instructions are generated in real time through the Actor network, and the trade-off between temperature field stability and energy consumption is evaluated through the Critic network. For the discrete action requirements of the slitting process, the Q-learning algorithm (Q-learning algorithm) can be used. The ε-greedy strategy is used to find the optimal solution in a short time. The triple specificity algorithm of the PSO algorithm, the DDPG algorithm and the Q-learning algorithm is used to solve the limitations of the traditional single algorithm for the different types of processes. PSO can prevent the coating parameters from falling into the local optimum, while DDPG can solve the time-delay control problem of the oven. Q-learning realizes efficient search of discrete slitting decisions, providing high-quality strategy initial values ​​for subsequent processing.

[0049] It is understandable that the PSO algorithm can solve high-dimensional continuous parameter optimization problems by simulating the collaborative foraging behavior of bird flocks. Each particle represents a candidate solution, such as the combination of coating gap and pressure. By tracking the individual historical optimal and group optimal update position, the speed update depends on the inertia term, that is, the early exploration, the individual cognitive term, that is, learning from its own optimality and the social cognitive term, that is, learning from the group optimality. The position iteration approaches the global optimality. Secondly, the DDPG algorithm integrates policy gradient and Q learning to solve the temporal decision-making problem in the continuous action space. The DDPG algorithm has an Actor-Critic dual network architecture. The ctor network (strategy network) generates continuous actions based on states, such as temperature field gradients, such as fan frequency conversion values. The critic network (value network) evaluates the long-term benefits of actions and achieves progressive optimization of strategies through soft updates of the target network and noise exploration of the OU process. Q-learning, based on temporal difference reinforcement learning, optimizes discrete action decision-making problems. It maintains a Q table to record state-action values, such as the expected benefit of the "medium speed + medium tension" combination of a slitting machine. It balances exploration and utilization through an ε-greedy strategy, such as randomly trying new actions 30% of the time. The Q value update integrates immediate rewards and future discounted benefits.

[0050] Step S2000-3: Perform secondary optimization on the original production process control strategy based on the Nash Q-learning algorithm to obtain a pending production process optimization strategy, obtain physical constraint data, verify the pending production process optimization strategy based on the physical constraint data, and after the verification is completed, obtain the production process optimization strategy.

[0051] Specifically, after generating the initial production process control original strategy, the utility value of each combination in the production process optimization strategy is calculated based on the Nash Q-learning algorithm, and then the Boltzmann exploration equilibrium strategy is updated. The Nash equilibrium point is determined by the convergence of the Q value to ensure that unilateral strategy deviation will result in loss of utility value, and solve the multi-objective conflict problem in the production process optimization strategy, such as the problem of improving coating quality and preventing tape breakage in slitting, to achieve global optimization.

[0052] It can be understood that the Nash Q-learning algorithm solves the multi-agent game problem by extending the traditional Q-learning framework, by upgrading the state-action value function Q of a single agent to the Nash Q-function under the joint action space, where each agent evaluates the expected cumulative reward of the joint action of all agents under the state; the algorithm iteratively updates the Q-value, replacing the maximum utility of traditional Q-learning with the Nash equilibrium utility, that is, the Q-value corresponding to the Nash equilibrium solution of the next state instead of max Q, and adopts the Boltzmann exploration strategy to balance exploration and utilization, and finally converges to the Nash equilibrium point. At this time, any agent unilaterally deviates from the strategy will reduce its own benefits, thereby coordinating the goal conflicts of the agents in the coating, curing, slitting and other process steps in the production of light-shielding tape, and realizing multi-party collaborative optimization.

[0053] Furthermore, step S2000-3 specifically includes the following steps: Step S2000-3-1: Obtain physical constraint conditions based on physical constraint data.

[0054] Specifically, physical constraint data can be extracted based on the production process association map, that is, relevant data during data cleaning, to prevent the parameters in the generated strategy from exceeding the physical limit control parameters of process equipment or process.

[0055] Step S2000-3-2: If the pending production process optimization strategy satisfies the physical constraint conditions, redefine the pending production process optimization strategy as the original production process control strategy.

[0056] Step S2000-3-4: If the pending production process optimization strategy does not satisfy the physical constraint conditions, the pending production process optimization strategy is modified until the physical constraint conditions are satisfied, and the original production process control strategy is obtained.

[0057] Specifically, corrections are made to the types of production process optimization strategies that do not meet the physical constraints. These may include exceeding the thermal safety boundary, exceeding the physical limits of the equipment, and power system constraints. Corrections are required for the above-mentioned situations. For example, if the physical limits of the equipment are exceeded, a safety clamp correction can be used, that is, forced corrections are made to the exceeded parameters. For example, if the pressure limit is 1.2MPa, but the parameter is 1.25MPa, 1.25MPa is adjusted to 1.2MPa.

[0058] Step S2003: obtaining a real-time production process association set, defining a production utility objective function based on the real-time production process association set and the production process optimization strategy, and obtaining a multidimensional utility vector based on the production utility objective function.

[0059] Step S2004: executing a collaborative control algorithm based on the multidimensional utility vector to obtain a multidimensional collaborative utility vector, adjusting the production utility objective function based on the multidimensional collaborative utility vector, and obtaining a production process optimization control strategy.

[0060] Furthermore, obtaining a multidimensional utility vector based on the production utility objective function specifically includes the following steps: Step S2004 - 1 : obtaining first utility data, second utility data, and third utility data based on the production utility objective function, and generating a multidimensional utility vector based on the first utility data, the second utility data, and the third utility data.

[0061] Specifically, the real-time production process association set may include data such as real-time yield, real-time power, and real-time output rate. Yield utility modeling can be performed based on the real-time yield, energy consumption utility modeling can be performed based on the real-time power, and production capacity utility modeling can be performed based on the real-time output rate. During the modeling process, the modeling needs to be normalized to establish a suitable production utility objective function. The first utility data is the yield weight of the production utility objective function, the second utility data is the energy consumption utility weight, and the third utility data is the production capacity utility weight. The yield weight, energy consumption utility weight, and production capacity utility weight are combined into a multi-dimensional utility vector.

[0062] Step S2004-2: Import the multidimensional utility vector into the GA algorithm, generate an initial utility population based on the GA algorithm, perform mutation and crossover operations on the initial utility population to obtain a secondary utility population, screen the secondary utility population based on the tournament strategy, and after the screening is completed, output the pending multidimensional utility vector based on the GA algorithm.

[0063] Specifically, before executing the GA (genetic algorithm), the GA algorithm needs to be configured and initialized. For example, energy consumption, production, and good product are defined as leading factors. The target collaborative value is quantified through the dynamic fitness functions of these factors. For example, in the energy consumption-leading mode, the energy consumption utility is weighted X times to guide the algorithm toward energy-saving solutions, or when the quality is insufficient, the lowest yield utility is triggered to be X to guide the algorithm to maintain the yield rate bottom line. The secondary utility population is generated through SBX crossover and Gaussian mutation. Through a tournament, for example, the top 30% of the secondary utility population is retained as the secondary utility population. After iterative evolution, the final undetermined multidimensional utility vector is generated.

[0064] It can be understood that genetic algorithms simulate the biological evolution mechanism by decoding the optimization problem into chromosomes, constructing the initial population, and then cyclically executing selection to retain individuals with high fitness, crossover, that is, exchanging chromosome fragments to generate new solutions, and mutation, that is, random perturbation to prevent premature operations, so that the population can evolve from generation to generation to approach the global optimal solution. It can solve high-dimensional, nonlinear, and multi-peak complex optimization problems, such as multi-objective coordination of process parameters, breaking through the limitations of traditional gradient methods that are prone to falling into local optimality, and dynamically balancing the conflicting goals of yield, energy consumption, and production capacity in the production of light-shielding tapes to achieve efficient search for the optimal solution set.

[0065] Step S2004-3: extract N groups of data from the secondary utility population as a strategy candidate set, import the strategy candidate set and the undetermined multidimensional utility vector into the NSGA-II algorithm, optimize the undetermined multidimensional utility vector based on the NSGA-II algorithm, and obtain a multidimensional utility vector.

[0066] Specifically, the NSGA-II algorithm is used to perform secondary optimization on the predetermined multidimensional utility vector under the conflicting goals of quality, energy consumption, and production capacity to make it more balanced. First, a three-dimensional objective function is constructed to quantify the utility of the strategy. N groups of candidate solutions are stratified using non-dominated sorting. The first frontier layer contains strategies that are not dominated by any solution. The distribution density of solutions in the target space is evaluated through congestion calculation. Solutions in sparse areas are retained to maintain diversity, and finally a multidimensional utility vector is output.

[0067] It can be understood that the GA algorithm compresses multiple objectives into a single-objective fitness function, essentially looking for a single optimal solution under preset preferences, while the NSGA-II algorithm retains all non-dominated solutions through non-dominated sorting. If solution A and solution B do not dominate each other, they belong to the first frontier layer. The diversity of the solution set is maintained by crowding calculation, and finally a complete Pareto optimal solution set is output. The NSGA-II algorithm is continued to be executed on the basis of the GA algorithm to generate a more schemed and optimized multi-dimensional utility vector. The production utility objective function is adjusted with this multi-dimensional utility vector, so that the production process optimization control strategy generated by the production utility objective function is more diverse on the premise of optimizing parameters, thereby providing better reference data for subsequent digital twin verification.

[0068] Step S300: Perform digital twin verification on the production process optimization control strategy, generate a final production process control strategy, and perform real-time regulation based on the final production process control strategy.

[0069] Furthermore, step S300 specifically includes the following steps: Step S3000: construct a virtual production process condition network, import the production process optimization control strategy into the virtual production process condition network, and generate a virtual condition data set based on the virtual production process condition network.

[0070] Specifically, the production process condition network is established based on the GAN network and includes a generator, a discriminator and a loss function.

[0071] Specifically, the GAN network (Generative Adversarial Network) learns the distribution of real data through the game between the generator and the discriminator, constructs a virtual working condition data set covering extreme disturbances and low-probability risks through the GAN network, takes the production process optimization control strategy as input, and uses a combined architecture of the generator, discriminator and loss function. The generator adopts the ResNet-18 network to learn the disturbance distribution law of the production process working conditions, and the discriminator adopts the CNN network to improve the physical rationality of the generated data. The loss function is designed and determined based on the actual situation to improve the training stability of the GAN network, synthesize various industrial disturbances, including but not limited to sudden changes in temperature and humidity, sudden increases in vibration, etc., and add physical rules when synthesizing industrial disturbances to prevent the generated data from exceeding the physical limits, thereby outputting multiple sets of virtual working condition data in time and space, providing a richer scenario library for subsequent strategy verification than directly using historical data.

[0072] Step S3001: construct a production process condition prediction network, import the virtual condition data set and the production process optimization control strategy into the production process condition prediction network, and obtain a strategy-quality prediction mapping table based on the production process condition prediction network.

[0073] Specifically, the production process condition prediction network is established based on the LSTM network and includes dual channels, one set of dual channels is a process channel, and the other set of dual channels is a quality channel. The process channel is used to predict time series changes such as thickness, temperature, and tension, while the quality channel is used to predict defect changes such as warping probability and transmittance.

[0074] Specifically, the LSTM network uses a dual-channel mechanism to quantify the risks of the control strategy based on the virtual working condition dataset generated in step S3000, realizing the function from passive monitoring to active defense. First, the virtual working condition dataset and the production process optimization control strategy are input into the LSTM network. The LSTM-based encoder captures the cumulative effect of timing disturbances, such as a temperature rise rate that is too fast may cause the cross-linking degree of the glue layer to decrease. Then, the spatial distribution is analyzed and multi-dimensional defect probabilities are output, such as the edge warping probability of 0.18 and the bubble generation probability of 0.35. Further, through counterfactual intervention analysis, a predicted probability mapping table of the strategy and quality in the production process optimization control strategy is output.

[0075] Step S3002: Obtain historical operating condition defect probabilities, construct a production process operating condition entropy value calculation model based on the historical operating condition defect probabilities and the strategy-quality prediction mapping table, and obtain a strategy risk entropy data set based on the production process operating condition entropy value calculation model.

[0076] Specifically, a unified decision-making scale across defect types is constructed through information entropy, and traditional experience-driven fuzzy judgments are converted into computable mathematical indicators. First, a risk entropy model is established through KL divergence, and the defect probability distribution under historical normal working conditions is used as a benchmark, such as edge warping of 5%, light transmittance exceeding the standard by 3%, and bubble generation of 5%. The deviation value of the probability of occurrence of each defect under the current strategic working condition is calculated. For example, a temperature rise of 4 degrees Celsius may cause the transmittance to increase by 7%. An adaptive learning mechanism is added. For example, after the production of every 200 rolls of light-shielding tape, the benchmark distribution is updated to enable the model to adapt to changes in working conditions in real time. The entropy weight in the production process condition entropy value calculation model is defined in combination with the strategy-quality prediction mapping table, thereby establishing a production process condition entropy value calculation model, and generating a strategic risk entropy data set through the production process condition entropy value calculation model.

[0077] Step S3003: Establish a control strategy decision surface, import the strategy risk entropy data set, strategy-quality prediction mapping table and production process optimization control strategy into the control strategy decision surface, and generate the final production process control strategy based on the control strategy decision surface.

[0078] Specifically, the control strategy decision surface is established based on the B-spline surface.

[0079] Specifically, by establishing a control strategy decision surface, the abstract multi-objective optimization is converted into an interactive geometric surface, thereby generating the final real-time control strategy, and based on the final real-time control strategy, the production process of the shading tape is optimized and adjusted in real time. First, the strategy-quality prediction mapping table is used as the X-axis, the strategy risk entropy data set is used as the Y-axis, and the production process optimization control strategy is used as the Z-axis. A continuous decision space is generated by fitting a third-order B-spline surface, the process constraints are dynamically mapped through the surface morphology, and the optimal path is automatically extracted based on the gradient ascent algorithm, thereby generating the final generation process control strategy. Finally, the final generation process control strategy is encapsulated as control instructions and sent to each production equipment in the process.

[0080] Step S4000: Acquire the post-regulation correlation data, and update the production process correlation map based on the post-regulation correlation data.

[0081] In some possible embodiments, after the final control is completed, a step of updating the production process association map generated in step S100 may be added.

[0082] In some possible embodiments, the production process association map can be updated and reconstructed through methods such as counterfactual reasoning and dynamic knowledge reconstruction. For example, the actual yield data and the real-time production process control set, that is, the associated data after regulation, can be obtained by performing an in-depth analysis of the deviation data, and analyzed by relevant algorithms of time series mining. For example, the Matrix Profile algorithm can be used to locate the maximum deviation time window by calculating the dynamic time-regularized distance between the actual and yield rates, and output a deviation event report based on, for example, a Bayesian diagnostic network. After that, reasoning is performed through the random forest algorithm to output a causal reasoning diagnostic report. Finally, the diagnostic report and the production process association map are reasoned and analyzed to obtain an updated production process association map.

[0083] Figure 3 A schematic diagram of a production process optimization control system for a light-shielding tape provided by some embodiments of the present application and capable of realizing the concept of the present application is shown. The following is a detailed introduction to the production process optimization control method for the light-shielding tape.

[0084] Specifically, a production process optimization control system for light-shielding tape includes: The atlas module is used to obtain a production process related data set, extract key information in the production process related data set, and construct a production process related atlas.

[0085] A primary strategy module, which is used to obtain a real-time production process control set, construct a production process agent cluster of the real-time production process control set and a production process association map, and generate a production process optimization control strategy through a virtual optimization strategy of the agent cluster; A secondary strategy module, which is used to perform digital twin verification and generate a final control strategy for the production process; A control module, which performs real-time control based on the final control strategy of the production process; The reconstruction module is used to obtain the associated data after regulation and update the production process association map.

[0086] The specific usage and function of this embodiment are described below: This method first obtains the production-related data set and extracts key information from the production process-related data set to construct a production process-related map, providing low-dimensional and interpretable physical constraint input for the processing of subsequent steps. Secondly, it obtains the real-time production process control set and constructs a production process agent cluster of the real-time production process control set and the production process-related map. The combined targeted algorithm of the PSO algorithm, the DDPG algorithm and the Q-learing algorithm can be used to perform targeted processing on different types of observation state data in the production process agent cluster to provide initial strategy initial values. Afterwards, the production process optimization control strategy is verified by digital twin to generate a production process agent cluster. The final control strategy of the process is obtained, thereby eliminating the low-quality strategies in the initial strategies, and based on this, the optimal control strategy is obtained. Finally, the production process association map can be updated based on the associated data after regulation to achieve global control of the process of light-shielding tape, and avoid the situation where local deterioration is caused by the traditional single-process parameter tuning. In this way, it adapts to the coupling effect, and through digital twin verification, the risky strategies that may appear in the initial strategy can be comprehensively excluded, so as to warn of compound risks and ensure that the optimal control strategy finally output has global control of the process and a control strategy with risk assurance, thereby providing a production process optimization control method and system for light-shielding tape.

[0087] Figure 4 A schematic diagram of an electronic device provided by some embodiments of the present application and capable of realizing the concept of the present application is shown, and the electronic device is introduced in detail below.

[0088] Specifically, an electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method proposed in the first embodiment of the present invention.

[0089] The following is a detailed introduction to the various components of electronic equipment: The term "processor" is the control center of an electronic device and can be a single processor or a collective term for multiple processing elements. For example, the processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the first embodiment of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0090] The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0091] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0092] The memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and be coupled to the processor via an interface circuit of the electronic device, and this is not specifically limited in the embodiments of the present invention.

[0093] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wireless communication (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0094] It should be understood that the term "and / or" as used herein simply describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent the existence of A alone, the existence of both A and B, or the existence of B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the related objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0095] It should be understood that in the embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0096] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A production process optimization control method for light-shielding tape, characterized in that: The method comprises: S100, obtaining a production process related data set, extracting key information from the production process related data set, and establishing a production process related map; S200, obtaining a real-time production process control set, building a production process agent cluster based on the real-time production process control set and the production process association map, and generating a production process optimization control strategy through an agent cluster virtual optimization strategy; S300, performing digital twin verification on the production process optimization control strategy, generating a final production process control strategy, and performing real-time regulation based on the final production process control strategy; S400: Acquire post-control correlation data, and update the production process correlation map based on the post-control correlation data.

2. The method for optimizing and controlling the production process of a light-shielding tape according to claim 1, characterized in that: Acquiring a production process related data set, extracting key information from the production process related data set, and establishing a production process related map, including: acquiring a production process related data set based on the production monitoring equipment group, and performing data processing operations on the production process related data set; Extracting key indicators of production processes from a production process-related data set through a neural network, and establishing a key production process matrix based on the key indicators of production processes; Calculating the production linear weight, production nonlinear weight, and production time lag weight based on the production process key matrix, importing the production process key matrix into the PC algorithm, and generating a basic production process association map based on the PC algorithm; The production process association map is generated by the basic production process association map, the production linear weight, the production nonlinear weight and the production lag weight.

3. The method for optimizing and controlling the production process of a light-shielding tape according to claim 1, characterized in that: Obtain a real-time production process control set, build a production process agent cluster based on the real-time production process control set and the production process association map, and generate a production process optimization control strategy through the agent cluster virtual optimization strategy, including: Initialize the agent environment based on the real-time production process control set and the production process association map to generate a production process agent cluster; Acquire process agent observation data based on the production process agent cluster, process the process agent observation data based on a triple specificity algorithm, and acquire an original production process control strategy; Performing secondary optimization on the original production process control strategy to obtain a production process optimization strategy; Acquire a real-time production process association set, define a production utility objective function based on the real-time production process association set and the production process optimization strategy, and acquire a multidimensional utility vector based on the production utility objective function; A collaborative control algorithm is executed based on the multidimensional utility vector to obtain a multidimensional collaborative utility vector, and the production utility objective function is adjusted based on the multidimensional collaborative utility vector to obtain a production process optimization control strategy.

4. The method for optimizing and controlling the production process of a light-shielding tape according to claim 3, characterized in that: Obtain production process optimization strategies, including: Based on the real-time production process control set and the production process association map, a first production process agent, a second production process agent, and a third production process agent are defined and combined into the production process agent cluster; based on the production process agent cluster, observation state data of the first agent, observation state data of the second agent, and observation data of the third agent are obtained; Import the observation state data of the first agent into the PSO algorithm, import the observation state data of the second agent into the DDPG algorithm, and import the observation data of the third agent into the Q-learning algorithm to obtain the original production process control strategy; The original production process control strategy is optimized secondary based on the Nash Q-learning algorithm to obtain a pending production process optimization strategy, obtain physical constraint data, and verify the pending production process optimization strategy based on the physical constraint data. After the verification is completed, the production process optimization strategy is obtained.

5. The method for optimizing and controlling the production process of a light-shielding tape according to claim 4, characterized in that: The undetermined production process optimization strategy is verified based on the physical constraint data. After the verification is completed, the original production process control strategy is obtained, including: Obtaining physical constraint conditions based on physical constraint data; If the undetermined production process optimization strategy satisfies the physical constraint condition, redefining the undetermined production process optimization strategy as the production process control original strategy; If the pending production process optimization strategy does not satisfy the physical constraint conditions, the pending production process optimization strategy is modified until the physical constraint conditions are satisfied, and the original production process control strategy is obtained.

6. The method for optimizing and controlling the production process of a light-shielding tape according to claim 3, characterized in that: Obtaining a multidimensional utility vector based on the production utility objective function includes: Acquiring first utility data, second utility data, and third utility data based on the production utility objective function, and generating a multidimensional utility vector based on the first utility data, the second utility data, and the third utility data; Importing the multidimensional utility vector into a GA algorithm, generating an initial utility population based on the GA algorithm, performing mutation and crossover operations on the initial utility population to obtain a secondary utility population, screening the secondary utility population based on a tournament strategy, and after the screening is completed, outputting a pending multidimensional utility vector based on the GA algorithm; N groups of data are extracted from the quadratic utility population as a strategy candidate set. The strategy candidate set and the undetermined multidimensional utility vector are introduced into the NSGA-II algorithm. The undetermined multidimensional utility vector is optimized based on the NSGA-II algorithm to obtain a multidimensional utility vector.

7. The method for optimizing and controlling the production process of a light-shielding tape according to claim 1, characterized in that: Perform digital twin verification on the production process optimization control strategy to generate the final production process control strategy, including: Constructing a virtual production process condition network, importing the production process optimization control strategy into the virtual production process condition network, and generating a virtual condition data set based on the virtual production process condition network; Constructing a production process condition prediction network, importing the virtual condition data set and the production process optimization control strategy into the production process condition prediction network, and obtaining a strategy-quality prediction mapping table based on the production process condition prediction network; Obtain historical working condition defect probabilities, construct a production process working condition entropy value calculation model based on the historical working condition defect probabilities and a strategy-quality prediction mapping table, and obtain a strategy risk entropy data set based on the production process working condition entropy value calculation model; A control strategy decision surface is established, and the strategy risk entropy data set, strategy-quality prediction mapping table and production process optimization control strategy are imported into the control strategy decision surface, and the final control strategy of the production process is generated based on the control strategy decision surface.

8. The method for optimizing and controlling the production process of a light-shielding tape according to claim 7, characterized in that: The method further comprises: The production process condition network is established based on the GAN network and includes a generator, a discriminator and a loss function. The generator is constructed by the ResNet-18 network, and the discriminator is constructed by the GAN network. The production process condition prediction network is established based on the LSTM network and includes dual channels, one set of dual channels is a process channel, and the other set of dual channels is a quality channel; The control strategy decision surface is established based on the B-spline surface.

9. A production process optimization control system for light-shielding tape, characterized in that: The system comprises: A graph module, which is used to obtain a production process related data set, extract key information from the production process related data set, and construct a production process related graph; A primary strategy module, which is used to obtain a real-time production process control set, construct a production process agent cluster of the real-time production process control set and a production process association map, and generate a production process optimization control strategy through a virtual optimization strategy of the agent cluster; A secondary strategy module, which is used to perform digital twin verification and generate a final control strategy for the production process; A control module, which performs real-time control based on the final control strategy of the production process; The reconstruction module is used to obtain the associated data after regulation and update the production process association map.

Images