Method and apparatus for color uniformity optimization for coating

Abstract

The present application relates to the field of coating technology, and provides a color uniformity optimization method and device for coating, the method comprising: initializing a preset number of decision process parameters randomly according to the constraint range of each decision process parameter, each group of decision process parameters being an individual, and obtaining an initial population; inputting each individual in the initial population and the current machine feature parameter combination into a color prediction model to obtain a color value uniformity index determined based on the prediction result of the color prediction model; and iteratively optimizing the initial population with the uniformity index as the fitness, and outputting each decision process parameter corresponding to the individual with the optimal fitness in the final population to the coating machine when the stop iteration condition is met. The present application obtains an accurate color value uniformity index through forward prediction of the color prediction model, and then obtains each decision process parameter with a globally optimal color value uniformity index through reverse optimization algorithm optimization of the decision process parameter value combination with the color value uniformity index as the target.

Description

Technical Field

[0001] This invention relates to the field of coating technology, and in particular to a method and apparatus for optimizing the color uniformity of coatings. Background Technology

[0002] HB (Humboldt) coating is a coating process designed to enhance the surface hardness of cover glass (CG). In the HB coating process, the uniformity of coating color is a core technical indicator that determines product performance, yield and reliability.

[0003] In existing coating processes, the optimization of coating color uniformity is mainly achieved by manually adjusting the process parameters of the coating machine, or by using an end-to-end model that predicts process parameters based on machine characteristics and color uniformity index data to output optimized process parameters. Manual machine adjustment relies heavily on personal experience. When multiple nonlinear, strongly coupled process parameters exist, and the coating machine exhibits state characteristic drift, manual adjustment requires repeated trial and error, resulting in long adjustment cycles and low optimization efficiency. Furthermore, manual adjustment is essentially a local search with a limited step size from the current center outwards, making it difficult to achieve the globally optimal solution for the combination of process parameters corresponding to the color uniformity target. In the other end-to-end model optimization approach, the model predicts process parameters based on machine characteristics and target quality data. The model relies on training data to cover the optimal solution, and the predicted process parameters are the average of historical process parameters, not the globally optimal solution for the combination of process parameters corresponding to the color uniformity target.

[0004] It is evident that neither manual machine adjustment nor model optimization can currently improve the accuracy of coating color uniformity control. Summary of the Invention

[0005] This invention provides a method and apparatus for optimizing the color uniformity of coatings, thereby solving the problem that existing optimization methods are unable to improve the accuracy of color uniformity control in coatings.

[0006] This invention provides a method for optimizing the color uniformity of a coating, comprising the following steps: According to the constraint range of each decision process parameter, a preset number of decision process parameters are randomly initialized, with each group of decision process parameters representing an individual, to obtain the initial population; The color prediction model is input after combining the feature parameters of each individual in the initial population and the current machine. The color value uniformity index is obtained based on the prediction results of the color prediction model. Using color value uniformity as the fitness index, the initial population is iteratively optimized. When the stopping condition is met, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine.

[0007] According to the present invention, a method for optimizing color uniformity in coating is provided, wherein the color prediction model is trained based on decision process parameter samples, machine feature parameter samples and corresponding color value vector labels. The color prediction model is input by combining the feature parameters of each individual in the initial population and the current machine, and the uniformity index of color values ​​is obtained based on the prediction results of the color prediction model, including: The initial population is combined with the current machine feature parameters and then input into the color prediction model to obtain the color value vector output by the color prediction model. The color value vector is the vector of the color value of each workpiece to be coated after coating in a coating batch of the coating machine. The color uniformity index is calculated based on the color value vector.

[0008] According to the color uniformity optimization method for coating provided by the present invention, after outputting the decision process parameters corresponding to the individual with the best fitness in the final population to the coating machine, the method further includes: The coating equipment is coated according to its characteristic parameters and the decision process parameters corresponding to the individual with the best fitness in the final population, and the color value vector detection value is obtained. The decision-making process parameters, machine feature parameters, and color value vector detection values ​​corresponding to the individual with the best fitness in the final population are stored as incremental samples in the training sample database. Incremental training of the color prediction model is triggered periodically based on incremental samples.

[0009] According to the present invention, a method for optimizing color uniformity in coating is provided, wherein the color prediction model is trained based on decision process parameter samples, non-decision process parameter samples, machine feature parameter samples and corresponding color value vector labels. The color prediction model is input by combining the feature parameters of each individual in the initial population with those of the current machine, resulting in a color value vector output by the model, including: Each individual in the initial population is combined with non-decision process parameters and current machine characteristic parameters and then input into the color prediction model to obtain the color value vector output by the color prediction model.

[0010] According to the present invention, a method for optimizing color uniformity in film coating is provided, which uses color value uniformity index as the fitness of the initial population for iterative optimization. When the stopping iteration condition is met, the method outputs the decision process parameters corresponding to the individual with the best fitness in the final population to the coating machine, including: Taking each individual in the initial population as the target individual, perform the following steps for any target individual: Mutation: Select the first, second, and third individuals that are different from any target individual, calculate the difference vector between the first and second individuals, scale the difference vector, and add it to the third individual to obtain the mutated individual; Crossover: Mix and cross the decision process parameters of the mutant individual with any target individual to generate the experimental individual. The experimental individual and the current machine feature parameters are combined and then input into the color prediction model. The fitness of the experimental individual is determined based on the prediction results of the color prediction model. Selection: Compare the fitness of the experimental individual with the fitness of any target individual, and select the individual with better fitness to enter the next generation of the population; Iterative convergence: Repeatedly execute mutation, crossover and selection steps until the preset number of generations or the optimal fitness of the population remains unchanged for multiple consecutive generations, thus satisfying the convergence condition; The decision-making process parameters corresponding to the individual with the best fitness in the final population after convergence are output to the coating machine.

[0011] According to the present invention, a method for optimizing the color uniformity of coating is provided, wherein the color value uniformity index includes the standard deviation or range of the color values ​​of each workpiece to be coated in a single coating batch of a coating machine.

[0012] According to the present invention, a method for optimizing the color uniformity of a coating is provided. For the HB coating process, the decision process parameters include the inert gas flow rate of the pores in the material film layer that has a direct impact on the color uniformity of the coating.

[0013] The present invention also provides a device for optimizing the color uniformity of a coating, comprising the following modules: The parameter initialization module is used to randomly initialize a preset number of decision process parameters according to the constraint range of each decision process parameter. Each group of decision process parameters is an individual, thus obtaining an initial population. The uniformity index determination module is used to input the combination of each individual in the initial population and the current machine feature parameters into the color prediction model to obtain the color value uniformity index determined based on the prediction results of the color prediction model. The iterative optimization module is used to iteratively optimize the initial population with the color value uniformity index as the fitness. When the stopping iteration condition is met, it outputs the decision process parameters corresponding to the individual with the best fitness in the final population to the coating machine.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the color uniformity optimization method for coating as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the color uniformity optimization method for coating as described above.

[0016] The color uniformity optimization method for coating provided by this invention inputs the combination of each individual in the population and the current machine's characteristic parameters into a color prediction model, obtaining a color value uniformity index determined based on the prediction results of the color prediction model. Through the model's precise characterization of the nonlinear and cross-coupling effects between parameters, an accurate color value uniformity index can be obtained. Then, by using the color value uniformity index as the objective and employing a reverse optimization algorithm to optimize the combination of decision process parameter values, the optimal decision process parameters corresponding to the individual with the best color value uniformity index can be obtained, i.e., the global optimal solution. This solves the shortcomings of manual adjustment in taking into account multiple parameter couplings and the difficulty of traditional models in obtaining a global optimal solution, significantly improving the control accuracy of coating color uniformity. Moreover, compared to manual optimization methods, while significantly improving the control accuracy of coating color uniformity, it also achieves an order-of-magnitude leap in machine adjustment efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts of the method for optimizing the color uniformity of film coating provided by the present invention.

[0019] Figure 2 This is the second schematic diagram of the process for optimizing the color uniformity of coating provided by the present invention.

[0020] Figure 3 This is a schematic diagram of the structure of the color uniformity optimization device for coating provided by the present invention.

[0021] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] The method for optimizing the color uniformity of film coating according to embodiments of the present invention, such as Figure 1 As shown, the procedure includes steps S110 to S130.

[0024] Step S110: Randomly initialize a preset number of decision process parameters according to the constraint range of each decision process parameter. Each group of decision process parameters represents one individual, resulting in an initial population. The preset number of groups can be two or more. The decision process parameters are process parameters that affect the coating color. In the initial population, the values ​​of the decision process parameters corresponding to different individuals will not be identical. The constraint range refers to the constraints on the values ​​of the decision process parameters during the coating process. For example, the decision process parameter can be the inert gas (e.g., argon) flow rate of one or more pores in the coating. Its constraint range can be process constraints such as the total inert gas flow rate of a certain pore layer not exceeding a set value, or the interlayer inert gas flow rate ratio. Of course, different coating processes may have different decision process parameters and constraint ranges.

[0025] Step S120: Input the combination of each individual in the initial population and the current machine feature parameters into the color prediction model to obtain the color value uniformity index determined based on the prediction results of the color prediction model. It can be understood that this color prediction model is trained based on samples of decision process parameters and machine feature parameters. Its prediction result can be a color value uniformity index directly, or it can be the color value of each workpiece to be coated in a single coating batch on the coating machine, and then the color value uniformity index is calculated based on the color value of each workpiece to be coated. For example, this color uniformity index can be the standard deviation or range of each color value; the smaller the standard deviation or range, the better the color uniformity.

[0026] The current equipment characteristic parameters are the characteristic parameters of the coating equipment acquired during the current coating process. For example, the characteristic parameters of the coating equipment may include: equipment number, batch number, equipment set rate, target material consumption, coating start pressure, evacuation time, sputtering source voltage, sputtering source current, sputtering source power, excitation source incident power, excitation source reflected power, sputtering source ARC (characterizing the argon content of the sputtering source), excitation source high-top oxygen flow rate, cavity vacuum gauge pressure, and cavity capacitor diaphragm pressure—at least one of these static and / or slowly variable parameters characterizing the current hardware state of the coating equipment. In this step, the current equipment characteristic parameters also serve as one of the inputs to the color prediction model, enabling the model to quantitatively perceive the drift of the coating equipment hardware state. When the hardware state of the coating equipment changes, the model can actively identify and perceive its impact on the coating result, thereby making the prediction result more consistent with the current hardware state of the coating equipment.

[0027] Step S130: Using the color value uniformity index as the fitness, iteratively optimize the initial population. When the stopping iteration condition is met, output the decision process parameters corresponding to the individual with the best fitness in the final population to the coating machine. That is, using the color value uniformity index as the objective, automatically search for the globally optimal combination of decision process parameters within the high-dimensional value space formed by the decision process parameters, thereby obtaining the fitness, i.e., the decision process parameters corresponding to the individual with the best color value uniformity index, thus completing one machine adjustment. These decision process parameters are then sent to the coating machine, which performs coating according to these parameters, improving the color uniformity of the coating. For example, the iterative optimization algorithm can be: differential evolution algorithm, genetic algorithm, particle swarm optimization algorithm, ant colony optimization algorithm, simulated annealing algorithm, Bayesian optimization algorithm, or gradient descent method, etc. When the color value uniformity index is the standard deviation or range of the color values, the individual with the best fitness in the final population is the individual with the smallest fitness, i.e., the individual with the smallest color value uniformity index.

[0028] In this embodiment, the method for optimizing the color uniformity of coating is used to input the combination of each individual in the population and the current machine's characteristic parameters into a color prediction model. This yields a color uniformity index determined by the model's prediction results. By accurately representing the nonlinear and cross-coupling effects between parameters through the model, an accurate color uniformity index can be obtained. Then, by using the color uniformity index as the objective and employing a reverse optimization algorithm to optimize the combination of decision process parameters, the optimal decision process parameters corresponding to the individual with the best color uniformity index can be obtained—the global optimal solution. This solves the shortcomings of manual adjustment, which cannot take into account the coupling of multiple parameters, and the difficulty of traditional models in obtaining a global optimal solution, significantly improving the control accuracy of coating color uniformity. Furthermore, compared to manual optimization, this method not only significantly improves the control accuracy of coating color uniformity but also achieves an order-of-magnitude leap in machine adjustment efficiency.

[0029] Furthermore, the color prediction model in this embodiment is a forward model, which predicts color results by deciding on process parameters and machine feature parameters. Compared to the traditional end-to-end model that predicts process parameters by machine features and target quality data (e.g., color target), the color prediction model in this embodiment, combined with backward optimization, has strong interpretability. The backward optimization process is transparent and convergence can be tracked. The traditional end-to-end model is a black box and it is difficult to explain why a certain process parameter is recommended. It can also add process parameter constraints more flexibly, while the traditional end-to-end model has difficulty handling complex constraints and usually needs to learn implicitly from data. Moreover, when the parameter space is large and the sample is insufficient, the forward color prediction model combined with backward optimization in this embodiment can utilize the generalization of the forward model to find the optimal combination that may not have appeared in the training data through global search. In contrast, the traditional end-to-end model relies on the training data to cover the optimal solution, and its optimal solution is only the average value of historical data, which cannot explore unknown better solutions.

[0030] In some embodiments, the color prediction model is trained based on decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels. That is, by training the color prediction model with decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels, an implicit mapping relationship is established between machine status, decision process parameters, and coating color quality, so that the trained color prediction model can predict color results and obtain color value uniformity index.

[0031] For example, the color prediction model can be an XGBoost model. During training, the sample dataset, including decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels, is divided into a training set and a validation set. Using a gradient boosting tree framework, multiple decision trees are iteratively generated with the mean squared error or mean absolute error as the loss function. Cross-validation is used to optimize key hyperparameters of XGBoost, such as the maximum tree depth, learning rate, subsampling ratio, and L1 / L2 regularization coefficients, to prevent overfitting and maximize prediction accuracy. After training, the model is run on an independent test set to predict model errors and verify whether its predicted color values ​​are highly consistent with the measured values.

[0032] Based on the trained color prediction model, in step S120, the characteristic parameters of each individual in the initial population and the current machine are combined and input into the color prediction model to obtain the color value uniformity index determined based on the prediction results of the color prediction model, specifically including: Step S121: Combine the features of each individual in the initial population with the current machine's feature parameters and input them into the color prediction model to obtain the color value vector output by the color prediction model. The color value vector is the vector of the color values ​​of each workpiece to be coated in one coating batch of the coating machine. For example, for HB coating, the color value vector is the vector of the color values ​​of a column of workpieces to be coated in one coating batch of the coating machine.

[0033] Step S122: Calculate the color uniformity index based on the color value vector. Specifically, calculate the color uniformity index based on each color value in the color value vector. For example, the color uniformity index is the standard deviation or range of each color value in the color value vector.

[0034] Understandably, color prediction models can also directly predict color uniformity indices. However, when predicting color uniformity indices directly from various parameters, the spatial distribution information of color values ​​is lost during the aggregation process of mapping each parameter to the color uniformity index. This makes it impossible to distinguish between two process anomalies: linear gradual drift (e.g., uneven airflow distribution at the top and bottom of the cavity due to blocked pores, resulting in regular color changes) and local center collapse (abnormal color of intermediate products due to fluctuations in the flow rate of certain middle-layer pores). These two phenomena may predict the same uniformity index, but their causes are completely different. Therefore, directly predicting the uniformity index cannot distinguish between these two extreme process anomalies, resulting in weak interpretability.

[0035] In this embodiment, the color prediction model does not directly predict the color uniformity index, but rather predicts the color value of each workpiece to be coated in a single coating batch of the coating machine. It can retain the relationship between a change in a certain parameter and the change in the color value of a certain workpiece to be coated. The model has strong interpretability and verifiability. Moreover, by predicting the color value of each workpiece to be coated, subsequent optimization can combine the physical correspondence between the position of the pores and the position of each product to accurately identify the causal relationship between adjusting which pore can improve which product, thereby achieving precise optimization and adjustment of decision-making process parameters. Furthermore, for HB coating, there is a clear longitudinal correspondence between the physical location of the pores and the location of the workpiece to be coated within the coating cavity: the airflow change of the i-th pore has the greatest impact on the adjacent products and the impact decreases on distant products. When the color prediction model outputs the color value of each piece, the influence weight of the flow rate of a certain pore on the color value of a certain piece can be visualized by analyzing the feature importance or gradient of the color prediction model. This not only makes the prediction results of the color prediction model interpretable and verifiable, but also allows the modeling to be guided by this prior local effect law when the amount of data is limited, thereby improving the generalization ability of the color prediction model.

[0036] In some embodiments, after outputting the decision-making process parameters corresponding to the individual with the best fitness in the final population to the coating machine in step S130, the method further includes: Step S140: Obtain the coating equipment's coating process parameters according to the equipment's characteristic parameters and the decision process parameters corresponding to the individual with the best fitness in the final population, and then detect the color value vector detection value. After receiving the decision process parameters corresponding to the individual with the best fitness in the final population, the coating equipment performs coating according to these decision process parameters. After coating, the color vector detection value is obtained through detection, and this color vector detection value can be used as a label for subsequent training data.

[0037] Step S150: Store the decision process parameters, machine feature parameters, and color value vector detection values ​​corresponding to the individual with the best fitness in the final population as incremental samples into the training sample database.

[0038] Step S160: Periodically trigger incremental training of the color prediction model based on incremental samples.

[0039] In this embodiment, the color vector detection data of each actual coating is stored in the training database, and the color prediction model is periodically triggered to perform incremental training. This ensures that the color prediction model's characterization of the mapping relationship between machine feature parameters, decision process parameters, and color results is always synchronized with the current real state of the coating machine, maintaining the accuracy and stability of the output color results over the long term. It also significantly reduces process fluctuations and unplanned downtime caused by machine state drift.

[0040] In some embodiments, the color prediction model is trained based on decision process parameter samples, non-decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels.

[0041] The process involves combining each individual in the initial population with the current machine feature parameters and inputting the combination into the color prediction model to obtain the color value vector output by the color prediction model. Specifically, this includes combining each individual in the initial population with the non-decision process parameters and the current machine feature parameters and inputting the combination into the color prediction model to obtain the color value vector output by the color prediction model.

[0042] Among them, the non-decision process parameter samples are process parameters that have little or no impact on color uniformity. For example, in an HB coating process with eleven layers, the decision process parameter is the inert gas flow rate of the pores during the coating of the fifth, sixth and eleventh layers. The non-decision process parameters may include: the speed coefficient of each coating layer, which determines the coating speed, and the inert gas flow rate of the pores during the coating of layers other than the fifth, sixth and eleventh layers.

[0043] In this embodiment, process parameters that have little or no impact on color uniformity (indirect impact means a small impact) are also used to train the color prediction model. During the training of the color prediction model, an implicit mapping relationship between all influencing factors and color values ​​is established, thereby making the color values ​​predicted by the trained color prediction model more accurate.

[0044] In some embodiments, step S130 involves iteratively optimizing the initial population using the color value uniformity index as the fitness metric. When the stopping iteration condition is met, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine. Specifically, the Differential Evolution (DE) algorithm is used to iteratively optimize the decision process parameters with the color value uniformity index as the objective. The algorithm process includes: Taking each individual in the initial population as the target individual, perform the following steps for any target individual: Mutation: Select a first, second, and third individual that are different from any target individual. Calculate the difference vector between the first and second individuals. Scale the difference vector and add it to the third individual to obtain the mutated individual. Specifically, for the target individual x... i Randomly select three different from x i individual x r1 x r2 and x r3 Calculate the first volume x r1 The second individual x r2 The difference vector x r1 -x r2 When scaling, a scaling factor F is introduced to obtain the mutated individual v. i =x r3 +F·(x r1 -x r2 F can take values ​​from 0.5 to 0.8 and is used to control the variable asynchronous length.

[0045] Crossover: The mutated individual is crossovered with any target individual using mixed decision-making process parameters to generate experimental individuals. These experimental individuals are then combined with the current machine's characteristic parameters and input into a color prediction model. The fitness of the experimental individual is determined based on the prediction results from the color prediction model. Specifically, the mutated individual v... i With any target individual x i Performing a binomial crossover, the result of the crossover along the j-th dimension for each individual is as follows: .

[0046] in, This represents an independent random number generated for the j-th dimension that is greater than 0 and less than 1, and CR represents the preset crossover probability. Used to specify that at least one dimension comes from the mutated individual v i , and These represent the mutated individual v. i and target individual x i The j-th dimension of the decision process parameters.

[0047] Selection: Compare the fitness of the experimental individual with the fitness of any target individual, and select the individual with better fitness to enter the next generation of the population. For example, when the standard deviation or mean of the color value is used as an indicator of color uniformity, i.e., fitness, the individual with lower fitness enters the next generation of the population.

[0048] Iterative convergence: The mutation, crossover and selection steps are executed repeatedly until the preset number of generations or the optimal fitness of the population remains unchanged for multiple consecutive generations, thus satisfying the convergence condition.

[0049] The decision-making process parameters corresponding to the individual with the best fitness in the final population after convergence are output to the coating machine. For example, when the standard deviation or mean of the color value is used as the color uniformity index, i.e., fitness, the individual with the lower fitness is the optimal individual.

[0050] Differential evolutionary algorithm utilizes the differences between individuals in the population to generate mutation directions. Combined with crossover and selection, it achieves global search to obtain the global optimum. This allows it to better capture the impact of changes in decision-making process parameters on the target during iterative optimization, and it can obtain a more accurate global optimum compared to other optimization algorithms. Moreover, individual evaluations in differential evolutionary algorithm are independent, and mutation, crossover, and selection operations can be performed in parallel for each individual in each generation of the population, improving the efficiency of iterative optimization.

[0051] In some embodiments, the color value uniformity index includes the standard deviation or range of the color values ​​of each workpiece to be coated in a single coating batch of the coating machine. The range is the difference between the maximum and minimum color values ​​of each workpiece to be coated in a single coating batch of the coating machine. A smaller standard deviation or range indicates more clustered data and better data consistency. Therefore, using the standard deviation or range of color values ​​as a color value uniformity index can better reflect the uniformity and consistency of color values.

[0052] In some embodiments, for the HB coating process, the decision process parameters include the inert gas flow rate of the pores in the material film layers that directly affect the color uniformity of the coating during coating. For example, in an HB coating process with eleven layers, the fifth layer is a silicon nitride material layer, the sixth layer is a silicon dioxide material layer, and the eleventh layer is a silicon oxynitride material layer. Experiments have shown that these three material layers have a direct impact on the uniformity of the coating color distribution, indicating a significant effect on the color uniformity of the coating. Therefore, for the eleven-layer HB coating process, the decision process parameters are the inert gas flow rates of the 10 pores (30 pores in total) distributed in each of the fifth, sixth, and eleventh layers during coating. During reverse optimization, only the inert gas flow rates corresponding to these three layers are iteratively adjusted. This results in fewer adjustment parameters, faster convergence, higher reverse optimization efficiency, and ultimately, a more uniform color value in the coating after adjusting the machine according to the optimized process parameters.

[0053] In some embodiments, taking the HB coating process as an example, the method flow for optimizing the color uniformity of the coating is as follows: Figure 2 As shown, the whole can be divided into two parts: one is the training and solidification of the color prediction model, and the other is the reverse optimization process based on the predicted values ​​of the color prediction model.

[0054] For the first part, the steps are as follows: Step S201: First, collect historical data. The historical data includes the decision process parameters, machine feature parameters, and color value vector of the workpiece to be coated after each coating process. This will give us the decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels.

[0055] Step S202: Perform data preprocessing and feature engineering on the current decision process parameter samples and machine feature parameter samples, including feature extraction and splicing of various types of parameters.

[0056] Step S203: Train a color prediction model using the features after data preprocessing and feature engineering, and the corresponding color value vector labels.

[0057] Step S204: During training, determine whether the model metrics meet the standards, i.e., whether the loss function of the predicted color value vector and the color value vector label has converged. If the loss function has converged, proceed to step S205; otherwise, proceed to step S202 to perform data preprocessing and feature engineering on the subsequent sample data.

[0058] Step S205: Solidify the color prediction model.

[0059] For the second part, the steps are as follows: Step S206: Initialize the decision process parameters of the preset number of groups, for example: in an HB coating process with eleven layers, the inert gas flow rate of 10 pores (30 pores in total) distributed in the fifth, sixth and eleventh layers of coating.

[0060] Step S207: Obtain the current machine characteristic parameters.

[0061] Step S208: Call the color prediction model to predict the color value vector of each individual in the population. That is, the color prediction model predicts the color value vector based on the currently input individual (i.e. a set of decision process parameters) and machine feature parameters.

[0062] Step S209: Calculate the color uniformity index, that is, calculate the color uniformity index based on the predicted color value vector, for example: calculate the range of color values ​​in the color value vector.

[0063] Step S210: Determine whether the iteration condition has been met. If yes, proceed to step S212; otherwise, proceed to step S211.

[0064] Step S211: Iterate and optimize the combination of decision process parameters using a reverse optimization algorithm, and then proceed to step S208, whereby the optimized decision process parameters and the current machine feature parameters are used again to predict the color value vector through a color prediction model. For example, for the differential evolution algorithm, the next generation of individuals and the current machine feature parameters are used again to predict the color value vector through the color prediction model.

[0065] Step S212: Output the combination of decision process parameter values ​​(inert gas values ​​for each layer). For HB coating, output the inert gas flow rate of each of the 10 pores (30 pores in total) distributed during the coating of the fifth, sixth and eleventh layers, and send the inert gas flow rate to the coating machine.

[0066] The color uniformity optimization apparatus for film coating provided by the present invention will be described below. The color uniformity optimization apparatus for film coating described below can be referred to in correspondence with the color uniformity optimization method for film coating described above.

[0067] The color uniformity optimization device for coating according to embodiments of the present invention, such as Figure 3 As shown, it includes the following modules: The parameter initialization module 310 is used to randomly initialize a preset number of decision process parameters according to the constraint range of each decision process parameter. Each group of decision process parameters is an individual, thus obtaining an initial population.

[0068] The uniformity index determination module 320 is used to input the combination of each individual in the initial population and the current machine feature parameters into the color prediction model to obtain the color value uniformity index determined based on the prediction results of the color prediction model.

[0069] The iterative optimization module 330 is used to iteratively optimize the initial population with the color value uniformity index as the fitness. When the stopping iteration condition is met, it outputs the decision process parameters corresponding to the individual with the best fitness in the final population to the coating machine.

[0070] In this embodiment, the color uniformity optimization device for coating is used to input the combination of each individual in the population and the current machine's characteristic parameters into a color prediction model. This yields a color uniformity index determined by the model's prediction results. By accurately representing the nonlinear and cross-coupling effects between parameters through the model, an accurate color uniformity index can be obtained. Then, by optimizing the combination of decision process parameters with the color uniformity index as the objective, the optimal decision process parameters corresponding to the individual with the best color uniformity index can be obtained—the global optimal solution. This solves the shortcomings of manual adjustment in taking into account multiple parameter couplings and the difficulty of traditional models in obtaining a global optimal solution, significantly improving the control accuracy of coating color uniformity. Furthermore, compared to manual optimization, it significantly improves the control accuracy of coating color uniformity while achieving an order-of-magnitude leap in machine adjustment efficiency.

[0071] In some embodiments, the color prediction model is trained based on decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels.

[0072] The uniformity index determination module 320 includes: The model prediction unit is used to input the combination of each individual in the initial population and the current machine feature parameters into the color prediction model to obtain the color value vector output by the color prediction model. The color value vector is the vector of the color value of each workpiece to be coated after coating in one coating batch of the coating machine.

[0073] The color uniformity index calculation unit is used to calculate the color uniformity index based on the color value vector.

[0074] In some embodiments, the color uniformity optimization device for coating further includes: The color value vector detection module is used to obtain the color value vector detection value after the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine. The coating machine applies coating according to the machine feature parameters and the decision process parameters corresponding to the individual with the best fitness in the final population.

[0075] The incremental sample update module is used to store the decision process parameters, machine feature parameters, and color value vector detection values ​​corresponding to the individual with the best fitness in the final population as incremental samples into the training sample database.

[0076] The incremental training trigger module is used to periodically trigger incremental training of the color prediction model based on incremental samples.

[0077] In some embodiments, the color prediction model is trained based on decision process parameter samples, non-decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels.

[0078] The model prediction unit is specifically used to input each individual in the initial population, along with non-decision process parameters and current machine characteristic parameters, into the color prediction model to obtain the color value vector output by the color prediction model.

[0079] In some embodiments, the iterative optimization module 330 is specifically used for: Taking each individual in the initial population as the target individual, perform the following steps for any target individual: Mutation: Select the first, second, and third individuals that are different from any target individual, calculate the difference vector between the first and second individuals, scale the difference vector, and add it to the third individual to obtain the mutated individual; Crossover: Mix and cross the decision process parameters of the mutant individual with any target individual to generate the experimental individual. The experimental individual and the current machine feature parameters are combined and then input into the color prediction model. The fitness of the experimental individual is determined based on the prediction results of the color prediction model. Selection: Compare the fitness of the experimental individual with the fitness of any target individual, and select the individual with better fitness to enter the next generation of the population; Iterative convergence: Repeatedly execute mutation, crossover and selection steps until the preset number of generations or the optimal fitness of the population remains unchanged for multiple consecutive generations, thus satisfying the convergence condition; The decision-making process parameters corresponding to the individual with the best fitness in the final population after convergence are output to the coating machine.

[0080] In some embodiments, the color value uniformity index includes the standard deviation or range of the color values ​​of each workpiece to be coated in a single coating batch of a coating machine.

[0081] In some embodiments, for the HB coating process, the decision process parameters include the inert gas flow rate of the material film layer during coating, which has a direct impact on the color uniformity of the coating.

[0082] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a method for optimizing the color uniformity of the coating, the method including: According to the constraint range of each decision process parameter, a preset number of decision process parameters are randomly initialized, with each group of decision process parameters representing an individual, thus obtaining the initial population.

[0083] The color prediction model is input by combining the characteristic parameters of each individual in the initial population and the current machine, and the color value uniformity index is obtained based on the prediction results of the color prediction model.

[0084] Using color value uniformity as the fitness index, the initial population is iteratively optimized. When the stopping condition is met, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine.

[0085] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the color uniformity optimization method for film coating provided by the above methods, the method comprising: According to the constraint range of each decision process parameter, a preset number of decision process parameters are randomly initialized, with each group of decision process parameters representing an individual, thus obtaining the initial population.

[0087] The color prediction model is input by combining the feature parameters of each individual in the initial population and the current machine, and the color value uniformity index is obtained based on the prediction results of the color prediction model.

[0088] Using color value uniformity as the fitness index, the initial population is iteratively optimized. When the stopping condition is met, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine.

[0089] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the color uniformity optimization method for film coating provided by the methods described above, the method comprising: According to the constraint range of each decision process parameter, a preset number of decision process parameters are randomly initialized, with each group of decision process parameters representing an individual, thus obtaining the initial population.

[0090] The color prediction model is input by combining the feature parameters of each individual in the initial population and the current machine, and the color value uniformity index is obtained based on the prediction results of the color prediction model.

[0091] Using color value uniformity as the fitness index, the initial population is iteratively optimized. When the stopping condition is met, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine.

[0092] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0093] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0094] All actions involving the acquisition of signal information or data in this invention are carried out in accordance with the relevant data protection laws and policies of the country where the device is located, and with the authorization granted by the owner of the device.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the color uniformity of a coating, characterized in that, include: According to the constraint range of each decision process parameter, a preset number of decision process parameters are randomly initialized, with each group of decision process parameters representing an individual, to obtain the initial population; The color prediction model is input after combining the feature parameters of each individual in the initial population and the current machine. The color value uniformity index is obtained based on the prediction results of the color prediction model. The initial population is iteratively optimized using the color value uniformity index as the fitness. When the iteration stops, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine.

2. The method for optimizing color uniformity in film coating according to claim 1, characterized in that, The color prediction model is trained based on decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels. The color prediction model is input by combining the characteristic parameters of each individual in the initial population and the current machine, and the color value uniformity index is obtained based on the prediction results of the color prediction model, including: The color prediction model is input after combining the individual in the initial population with the current machine feature parameters to obtain the color value vector output by the color prediction model. The color value vector is the vector of the color values ​​of each workpiece to be coated after coating in a coating batch of the coating machine. The color uniformity index is calculated based on the color value vector.

3. The method for optimizing color uniformity in film coating according to claim 2, characterized in that, After outputting the decision-making process parameters corresponding to the individual with the best fitness in the final population to the coating machine, the following steps are also included: The coating machine is coated according to the machine's characteristic parameters and the decision process parameters corresponding to the individual with the best fitness in the final population, and the color value vector detection value is obtained. The decision-making process parameters, machine feature parameters, and color value vector detection values ​​corresponding to the individual with the best fitness in the final population are stored as incremental samples in the training sample database. Incremental training of the color prediction model is periodically triggered based on the incremental samples.

4. The method for optimizing color uniformity in film coating according to claim 2, characterized in that, The color prediction model is trained based on decision process parameter samples, non-decision process parameter samples, machine feature parameter samples, and corresponding color value vector labels. The color prediction model is obtained by combining the feature parameters of each individual in the initial population and the current machine, and then inputting the combination into the color prediction model to obtain the color value vector output by the color prediction model, including: Each individual in the initial population is combined with the non-decision process parameters and the current machine characteristic parameters and then input into the color prediction model to obtain the color value vector output by the color prediction model.

5. The method for optimizing the color uniformity of a coating according to any one of claims 1 to 4, characterized in that, Using the color value uniformity index as the fitness iterative optimization initial population, when the iteration stopping condition is met, the decision process parameters corresponding to the individual with the best fitness in the final population are output to the coating machine, including: Taking each individual in the initial population as the target individual, perform the following steps for any target individual: Mutation: Select the first, second, and third individuals that are different from any target individual, calculate the difference vector between the first and second individuals, scale the difference vector, and add it to the third individual to obtain the mutated individual; Crossover: The mutant individual is crossovered with any target individual by mixing decision process parameters to generate a test individual. The test individual and the current machine feature parameters are combined and input into the color prediction model to obtain the fitness of the test individual based on the prediction results of the color prediction model. Selection: Compare the fitness of the experimental individual with the fitness of any target individual, and select the individual with better fitness to enter the next generation of the population; Iterative convergence: Repeatedly execute mutation, crossover and selection steps until the preset number of generations or the optimal fitness of the population remains unchanged for multiple consecutive generations, thus satisfying the convergence condition; The decision-making process parameters corresponding to the individual with the best fitness in the final population after convergence are output to the coating machine.

6. The method for optimizing color uniformity of a coating according to any one of claims 1 to 4, characterized in that, The color value uniformity index includes the standard deviation or range of the color values ​​of each workpiece to be coated in a single coating batch on the coating machine.

7. The method for optimizing the color uniformity of a coating according to any one of claims 1 to 4, characterized in that, For the HB coating process, the decision process parameters include the inert gas flow rate of the material film layer during coating, which has a direct impact on the color uniformity of the coating.

8. A device for optimizing the color uniformity of a coating, characterized in that, include: The parameter initialization module is used to randomly initialize a preset number of decision process parameters according to the constraint range of each decision process parameter. Each group of decision process parameters is an individual, thus obtaining an initial population. The uniformity index determination module is used to combine the characteristic parameters of each individual in the initial population and the current machine and input them into the color prediction model to obtain the color value uniformity index determined based on the prediction results of the color prediction model. The iterative optimization module is used to iteratively optimize the initial population with the color value uniformity index as the fitness. When the stopping iteration condition is met, it outputs the decision process parameters corresponding to the individual with the best fitness in the final population to the coating machine.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method for optimizing the color uniformity of coating as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the color uniformity optimization method for coating as described in any one of claims 1 to 7.

Images