Particle swarm optimization design method for coating parameters of heat-insulating glass

Abstract

The invention discloses a particle swarm optimization design method for heat insulation glass coating parameters, and belongs to the technical field of glass coating process optimization. The method comprises the following steps: collecting whole-process basic data related to heat-insulating glass coating, completing standardized pretreatment, and establishing a heat-insulating glass coating total-factor standardized database; coating process controllable parameters are selected to construct a multi-objective optimization total objective function, and constraint conditions of an optimization process are set in combination with working condition boundary data; based on the database, particle swarm algorithm population initialization is completed, and a global optimal coating process parameter combination is output through iterative optimization; and parameter correction is completed through simulation and test verification, and final filming process parameters capable of falling to the ground are output. According to the method, excessive dependence of artificial experience on process design can be avoided, global collaborative optimization of coating multi-process parameters is realized, the standardization degree of coating process design is improved, and the consistency of coating product performance and mass production suitability are improved.

Description

Technical Field

[0001] This invention relates to the field of glass coating process optimization technology, and in particular to a particle swarm optimization design method for thermal insulation glass coating parameters. Background Technology

[0002] Thermal insulation glass coating technology is one of the core technologies for achieving energy conservation and emission reduction in multiple fields such as construction, automobiles, and home appliances. By preparing functional thin films on the surface of glass substrates, selective control of different wavelength spectrums can be achieved, blocking infrared heat radiation while ensuring visible light transmittance. This is currently the mainstream technical path for improving the thermal insulation and energy-saving performance of glass. With the continuous increase in the demand for energy conservation and emission reduction across society, the application scenarios of thermal insulation coated glass are constantly expanding. The downstream market has put forward higher requirements for the optical performance, thermal insulation performance, environmental resistance performance, and batch consistency of coated glass. The design and optimization of the coating process directly determines the final product performance and mass production effectiveness of thermal insulation glass. Currently, the mainstream heat-insulating glass coating manufacturing processes in the industry include various physical and chemical coating methods such as magnetron sputtering coating and vacuum evaporation coating. The coating production process involves multiple types of controllable process parameters. The industry has carried out a lot of technical research and production practice around the design, debugging and optimization of coating processes, and has formed a variety of mature process debugging methods. At the same time, the application research of intelligent optimization algorithms in the field of industrial production parameter optimization continues to deepen, and the integration and application of swarm intelligence optimization algorithms with industrial manufacturing scenarios is gradually being implemented, providing corresponding theoretical and practical foundations for the intelligent and standardized design of coating process parameters.

[0003] Currently, the design of heat-insulating glass coating process parameters still heavily relies on the manual experience of senior operators on the production floor. A standardized, end-to-end data support system has not yet been established. Differences in experience and operating habits among different operators result in insufficient standardization in the process parameter design. Furthermore, data accumulated from historical production processes regarding materials, processes, performance, and operating conditions cannot be systematically integrated and effectively reused. This makes it difficult to fully incorporate comprehensive production-related data during the process parameter design, failing to provide comprehensive data support. Additionally, the optimization of existing coating process parameters often employs single-variable trial-and-error methods, making it difficult to achieve global collaborative optimization of multiple process parameters. This fails to address the optimization needs of multiple dimensions during coating production, often resulting in situations where one indicator meets the standard while others fail to meet usage requirements, making it difficult to obtain a well-suited combination of process parameters. Moreover, the existing process parameter design process lacks a complete simulation and experimental verification loop. The optimized process parameter combinations are not sufficiently adaptable to actual mass production scenarios, requiring repeated shutdowns and adjustments during mass production. This prevents the rapid output of coating process parameter combinations that can be directly applied, impacting the overall efficiency of coating production and the stability of product performance. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a particle swarm optimization design method for thermal insulation glass coating parameters.

[0005] The objective of this invention is achieved through the following technical solution: A particle swarm optimization design method for parameters of heat-insulating glass coating is provided, the method comprising the following steps: S1. Collect basic material characteristic data, process parameter data, performance test data and working condition boundary data related to heat insulation glass coating, preprocess all collected data, and build a standardized database of all elements of heat insulation glass coating according to the set classification rules. S2. Select controllable parameters of the coating process as optimization variables, construct a multi-objective optimization objective function, and combine the working condition boundary data in the standardized database of all elements of heat insulation glass coating to set the constraints of the optimization process; S3. Based on the standardized database of all elements of heat insulation glass coating, the population initialization of the particle swarm algorithm is completed. Combined with the multi-objective optimization total objective function and constraint conditions, iterative optimization is performed to update the individual optimal solution and the global optimal solution of the particle swarm. After the iteration terminates, the coating process parameter combination corresponding to the global optimal solution is output. S4. Perform simulation and experimental verification on the output coating process parameter combination, correct the parameters based on the verification results, and output the final feasible coating process parameter combination.

[0006] Furthermore, step S1 includes: S1.1. Collect basic material characteristic data, process parameter data, performance test data, and operating condition boundary data; basic material characteristic data includes substrate parameters, coating material parameters, and auxiliary material parameters; process parameter data includes controllable process parameters and auxiliary process parameters; performance test data includes film performance index data and production process index data; operating condition boundary data includes equipment rated parameter boundaries, standard mandatory requirement boundaries, and mass production environment fluctuation boundaries. S1.2. Perform outlier removal, missing value completion, and dimensionless normalization on all collected data to obtain standardized and valid data; S1.3. According to the three-level classification rule of substrate type, coating material system and application scenario, the effective data is classified and stored. The three-level classification rule is a hierarchical classification and storage rule with substrate type as the first level, coating material system as the second level and application scenario as the third level, and a standardized database of all elements of heat insulation glass coating is built. S1.4. After completing a single coating production, synchronize the actual production process parameters, performance test data and operating condition data to the standardized database of all elements of heat insulation glass coating, and complete the iterative update of the standardized database of all elements of heat insulation glass coating.

[0007] Furthermore, step S2 includes: S2.1. Select controllable parameters of the coating process as optimization variables from the controllable process parameters stored in the standardized database of all elements of heat insulation glass coating, and construct an optimization variable parameter vector; S2.2. Based on the three dimensions of membrane performance, production efficiency and production energy consumption, corresponding sub-objective functions are constructed respectively. A linear weighting method is used to combine the sub-objective functions. The linear weighting method is a processing method that assigns a corresponding weight coefficient to each sub-objective function and merges multiple sub-objective functions into a single objective function, thus constructing a multi-objective optimization overall objective function. S2.3. Based on the equipment rated parameter boundaries and standard mandatory requirement boundaries in the full-element standardized database of heat insulation glass coating, set hard constraints for the optimization process; S2.4. Based on the film performance index data and production process index data in the standardized database of all elements of heat insulation glass coating, set soft constraints for the optimization process.

[0008] Furthermore, step S3 includes: S3.1. Retrieve the historical process parameter combinations of the corresponding substrate type and coating material system from the standardized database of all elements of heat insulation glass coating, and complete the population initialization of particle swarm algorithm based on the retrieved parameter combinations, and assign initial position and initial velocity to each particle in the population. S3.2. Substitute the parameter vector corresponding to each particle into the overall objective function of the multi-objective optimization, and calculate the fitness value of each particle in combination with the constraints; S3.3. Compare the current fitness value of each particle with the historical best fitness value to update the individual optimal solution; traverse the individual optimal solutions of all particles in the population to update the global optimal solution; S3.4. Based on the adaptive inertia weight adjustment mechanism, update the velocity and position of each particle; the adaptive inertia weight adjustment mechanism is to dynamically adjust the inertia weight in the particle swarm velocity update formula according to the number of iterations during the iteration process. The upper limit weight value is set in the early stage of the iteration, and the lower limit weight value is set in the later stage of the iteration; detect the population diversity, which is characterized by the variance of the fitness values ​​of all particles in the population. When the population diversity is lower than the set threshold, perform mutation operation on some particles; S3.5. Determine if the iteration termination condition is met. If the termination condition is met, terminate the iteration and output the coating process parameter combination corresponding to the global optimal solution. If the termination condition is not met, return to the fitness value calculation step and continue the iteration.

[0009] Furthermore, step S4 includes: S4.1. Import the output coating process parameter combination into the optical thin film simulation software, complete the simulation calculation of the film optical performance, and verify whether the coating process parameter combination meets the constraints. S4.2. For the combination of coating process parameters that meets the constraints verified by simulation, perform laboratory coating pilot tests to complete the actual testing of film performance and production indicators; S4.3. Compare the small-scale test results with the constraints. If there is a deviation, synchronize the small-scale test data to the standardized database of all elements of heat-insulating glass coating and re-execute the iterative optimization. S4.4. For small-scale experiments that verify the coating process parameter combinations that meet the constraints, output the final, feasible coating process parameter combinations.

[0010] Furthermore, in step S1, anomaly filtering rules based on the normal distribution characteristics of the data are used to remove outliers from the collected data, eliminating invalid collected data that exceeds the set distribution range; multiple interpolation is used to complete missing values ​​in the collected data, filling in missing data in the fields that occur during the collection process; and linear transformation standardization is used to complete dimensionless processing of the collected data, uniformly mapping all collected data to a set numerical range, eliminating the interference of data with different dimensions on the subsequent optimization process.

[0011] Furthermore, in step S2, the sub-objective functions include a performance sub-objective function, a production efficiency sub-objective function, and a production energy consumption sub-objective function; the sum of the weight coefficients corresponding to the linear weighting method is 1, and the weight coefficients are adjusted according to the application scenario of the coated product; each sub-objective function corresponds to a set calculation logic, and the sub-objective functions are merged into a single multi-objective optimization overall objective function through the linear weighting method.

[0012] Furthermore, in step S3, the iteration termination conditions include two types: the first type is that the number of iterations reaches the set maximum number of iterations, and the second type is that the fitness value of the global optimal solution changes less than the set convergence accuracy threshold for multiple consecutive generations. If either of the two types of iteration termination conditions is met, the iterative optimization process is terminated, and the combination of coating process parameters corresponding to the global optimal solution obtained during the iteration process is output.

[0013] Furthermore, in step S4, the simulation verifies the coating process parameter combination that does not meet the constraints, adjusts the weight coefficients corresponding to the multi-objective optimization total objective function and the boundary range of the constraints, substitutes the adjusted multi-objective optimization total objective function and constraints into the particle swarm optimization algorithm, re-executes the complete iterative optimization process, outputs a new coating process parameter combination, and re-executes the simulation verification process.

[0014] Furthermore, after step S4 is completed, the final feasible coating process parameter combination is imported into the mass production coating line. Real-time data on operating condition fluctuations during the mass production process is collected. Based on the operating condition fluctuation data, micro-iterative optimization is performed on the coating process parameter combination. Micro-iterative optimization is an optimization process that keeps the main framework of the original coating process parameter combination unchanged and only performs small-scale iterative optimization on the parameter dimensions affected by operating condition fluctuations, thus completing the dynamic adjustment of parameters in the mass production process. The process parameters, performance test data, and operating condition data in the mass production process are synchronized to the standardized database of all elements of heat insulation glass coating, thus completing the continuous iterative update of the database.

[0015] The beneficial effects of this invention are: (1) Through the construction of a standardized database of all elements, the iterative optimization and verification of particle swarm algorithm, the global optimization of multiple parameters of the coating process is realized, the excessive reliance on manual experience in process design is eliminated, and the standardization of coating process parameter design is improved. (2) Build an optimization system that combines multi-dimensional sub-objective functions with dual constraints, take into account the synergistic optimization of membrane performance, production efficiency and production energy consumption, avoid the performance shortcomings of other dimensions caused by single-dimensional optimization, and ensure the comprehensive adaptability of process parameters. (3) By using the closed-loop verification and correction process of simulation and experiment, combined with the dynamic parameter adjustment mechanism of mass production conditions, the consistency and stability of the coating layer performance can be improved, the process debugging cycle can be shortened, and the raw material loss in the production process can be reduced. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the steps of a particle swarm optimization design method for parameters of heat-insulating glass coating. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 See Figure 1This embodiment provides a particle swarm optimization design method for thermal insulation glass coating parameters, which includes the following steps: S1. Collect basic material characteristic data, process parameter data, performance test data and working condition boundary data related to heat insulation glass coating, preprocess all collected data, and build a standardized database of all elements of heat insulation glass coating according to the set classification rules. S2. Select controllable parameters of the coating process as optimization variables, construct a multi-objective optimization objective function, and combine the working condition boundary data in the standardized database of all elements of heat insulation glass coating to set the constraints of the optimization process; S3. Based on the standardized database of all elements of heat insulation glass coating, the population initialization of the particle swarm algorithm is completed. Combined with the multi-objective optimization total objective function and constraint conditions, iterative optimization is performed to update the individual optimal solution and the global optimal solution of the particle swarm. After the iteration terminates, the coating process parameter combination corresponding to the global optimal solution is output. S4. Perform simulation and experimental verification on the output coating process parameter combination, correct the parameters based on the verification results, and output the final feasible coating process parameter combination.

[0019] In some embodiments, step S1 includes: S1.1. Collect basic material characteristic data, process parameter data, performance test data, and operating condition boundary data; basic material characteristic data includes substrate parameters, coating material parameters, and auxiliary material parameters; process parameter data includes controllable process parameters and auxiliary process parameters; performance test data includes film performance index data and production process index data; operating condition boundary data includes equipment rated parameter boundaries, standard mandatory requirement boundaries, and mass production environment fluctuation boundaries. S1.2. Perform outlier removal, missing value completion, and dimensionless normalization on all collected data to obtain standardized and valid data; S1.3. According to the three-level classification rule of substrate type, coating material system and application scenario, the effective data is classified and stored. The three-level classification rule is a hierarchical classification and storage rule with substrate type as the first level, coating material system as the second level and application scenario as the third level, and a standardized database of all elements of heat insulation glass coating is built. S1.4. After completing a single coating production, synchronize the actual production process parameters, performance test data and operating condition data to the standardized database of all elements of heat insulation glass coating, and complete the iterative update of the standardized database of all elements of heat insulation glass coating.

[0020] In some embodiments, step S2 includes: S2.1. Select controllable parameters of the coating process as optimization variables from the controllable process parameters stored in the standardized database of all elements of heat insulation glass coating, and construct an optimization variable parameter vector; S2.2. Based on the three dimensions of membrane performance, production efficiency and production energy consumption, corresponding sub-objective functions are constructed respectively. A linear weighting method is used to combine the sub-objective functions. The linear weighting method is a processing method that assigns a corresponding weight coefficient to each sub-objective function and merges multiple sub-objective functions into a single objective function, thus constructing a multi-objective optimization overall objective function. S2.3. Based on the equipment rated parameter boundaries and standard mandatory requirement boundaries in the full-element standardized database of heat insulation glass coating, set hard constraints for the optimization process; S2.4. Based on the film performance index data and production process index data in the standardized database of all elements of heat insulation glass coating, set soft constraints for the optimization process.

[0021] In some embodiments, step S3 includes: S3.1. Retrieve the historical process parameter combinations of the corresponding substrate type and coating material system from the standardized database of all elements of heat insulation glass coating, and complete the population initialization of particle swarm algorithm based on the retrieved parameter combinations, and assign initial position and initial velocity to each particle in the population. S3.2. Substitute the parameter vector corresponding to each particle into the overall objective function of the multi-objective optimization, and calculate the fitness value of each particle in combination with the constraints; S3.3. Compare the current fitness value of each particle with the historical best fitness value to update the individual optimal solution; traverse the individual optimal solutions of all particles in the population to update the global optimal solution; S3.4. Based on the adaptive inertia weight adjustment mechanism, update the velocity and position of each particle; the adaptive inertia weight adjustment mechanism is to dynamically adjust the inertia weight in the particle swarm velocity update formula according to the number of iterations during the iteration process. The upper limit weight value is set in the early stage of the iteration, and the lower limit weight value is set in the later stage of the iteration; detect the population diversity, which is characterized by the variance of the fitness values ​​of all particles in the population. When the population diversity is lower than the set threshold, perform mutation operation on some particles; S3.5. Determine if the iteration termination condition is met. If the termination condition is met, terminate the iteration and output the coating process parameter combination corresponding to the global optimal solution. If the termination condition is not met, return to the fitness value calculation step and continue the iteration.

[0022] In some embodiments, step S4 includes: S4.1. Import the output coating process parameter combination into the optical thin film simulation software, complete the simulation calculation of the film optical performance, and verify whether the coating process parameter combination meets the constraints. S4.2. For the combination of coating process parameters that meets the constraints verified by simulation, perform laboratory coating pilot tests to complete the actual testing of film performance and production indicators; S4.3. Compare the small-scale test results with the constraints. If there is a deviation, synchronize the small-scale test data to the standardized database of all elements of heat-insulating glass coating and re-execute the iterative optimization. S4.4. For small-scale experiments that verify the coating process parameter combinations that meet the constraints, output the final, feasible coating process parameter combinations.

[0023] In some embodiments, in step S1, anomaly filtering rules based on the normal distribution characteristics of the data are used to remove outliers from the collected data, and invalid collected data that exceeds the set distribution range is removed; multiple interpolation is used to complete missing values ​​of the collected data, and missing data in the fields that occur during the collection process is completed; linear transformation standardization is used to complete the dimensionless processing of the collected data, and all collected data are uniformly mapped to the set numerical range to eliminate the interference of data with different dimensions on the subsequent optimization process.

[0024] In some embodiments, in step S2, the sub-objective functions include a performance sub-objective function, a production efficiency sub-objective function, and a production energy consumption sub-objective function; the sum of the weight coefficients corresponding to the linear weighting method is 1, and the weight coefficients are adjusted according to the application scenario of the coating product; each sub-objective function corresponds to a set calculation logic, and the sub-objective functions are merged into a single multi-objective optimization overall objective function through the linear weighting method.

[0025] In some embodiments, in step S3, the iteration termination conditions include two types: the first type is that the number of iterations reaches the set maximum number of iterations, and the second type is that the fitness value of the global optimal solution changes less than the set convergence accuracy threshold for multiple consecutive generations; if either of the two types of iteration termination conditions is met, the iterative optimization process is terminated, and the coating process parameter combination corresponding to the global optimal solution obtained during the iteration process is output.

[0026] In some embodiments, in step S4, the simulation verifies the coating process parameter combination that does not meet the constraints, adjusts the weight coefficients corresponding to the multi-objective optimization total objective function and the boundary range of the constraints, substitutes the adjusted multi-objective optimization total objective function and constraints into the particle swarm optimization algorithm, re-executes the complete iterative optimization process, outputs a new coating process parameter combination, and re-executes the simulation verification process.

[0027] In some embodiments, after step S4 is completed, the final feasible coating process parameter combination is imported into the mass production coating line. Operating condition fluctuation data during the mass production process is collected in real time. Based on the operating condition fluctuation data, micro-iterative optimization is performed on the coating process parameter combination. Micro-iterative optimization is an optimization process that keeps the main framework of the original coating process parameter combination unchanged and only performs small-scale iterative optimization on the parameter dimensions affected by operating condition fluctuations, thereby completing the dynamic adjustment of parameters in the mass production process. The process parameters, performance test data and operating condition data in the mass production process are synchronized to the standardized database of all elements of heat insulation glass coating, thereby completing the continuous iterative update of the database.

[0028] Example 2 This embodiment presents a specific implementation process for a particle swarm optimization design method for heat-insulating glass coating parameters, fully covering the entire process of optimizing heat-insulating glass coating parameters. Through four stages—standardized data processing, construction of a multi-objective optimization system, iterative optimization using the particle swarm algorithm, and parameter verification and correction—the optimized design of the coating process parameters is completed. The heat-insulating glass coating involved in this embodiment refers to the existing process of depositing one or more thin films on the surface of a glass substrate using physical or chemical methods to achieve selective transmission of different spectral bands and thus achieve heat insulation. The particle swarm optimization algorithm involved is an existing stochastic optimization algorithm based on swarm intelligence, which completes optimization by simulating the cooperative behavior of biological groups during foraging. This embodiment combines this algorithm with the heat-insulating glass coating process for global optimization of multiple parameters in the coating process, solving the optimization problem under multi-parameter coupling. The complete implementation steps are as follows: S1. Construction of a comprehensive database for coating elements: S1.1. Data collection related to the entire coating process: This step involves collecting raw data related to the entire coating production process, providing a foundational data source for subsequent processing, database construction, and parameter optimization. All collected data corresponds to the same batch of coating production processes, ensuring accurate correspondence between different data types. The collected data includes four categories: basic material property data, process parameter data, performance test data, and operating condition boundary data.

[0029] Basic material property data includes substrate parameters, coating material parameters, and auxiliary material parameters. Substrate parameters include substrate type, thickness, visible light transmittance, coefficient of thermal expansion, and surface roughness, which directly affect the bonding effect between the film and the substrate and the optical performance of the finished product. Coating material parameters include the composition, purity, melting point, optical constants, and thermal conductivity of the coating material. Optical constants include refractive index and extinction coefficient, which together determine the spectral selectivity of the coating layer. The collected optical constants cover the working wavelength range corresponding to the coating process. Auxiliary material parameters include the composition and purity of sputtering gas and reactant gas, which affect the coating reaction rate and the quality of film formation.

[0030] The process parameter data includes controllable process parameters and auxiliary process parameters, which are the optimization objects in this embodiment. All combinations of process parameters used in historical production processes are collected to provide basic samples for subsequent algorithm optimization.

[0031] Performance testing data includes membrane performance indicators and production process indicators, serving as a direct basis for evaluating the quality of process parameter combinations. Membrane performance indicators include visible light transmittance, infrared blocking rate, ultraviolet blocking rate, membrane adhesion, acid and alkali corrosion resistance, weather resistance, and surface resistivity uniformity. All data are obtained according to standardized testing methods to ensure comparability between different batches. Production process indicators include single-batch commissioning cycle, production energy consumption, product yield, and equipment uptime, used to evaluate the mass production feasibility of the process parameter combinations.

[0032] Operating boundary data includes equipment rated parameter boundaries, standard mandatory requirement boundaries, and mass production environment fluctuation boundaries, which are constraints that must be followed during parameter optimization. Equipment rated parameter boundaries define the rated operating range of various parameters of the coating production equipment, ensuring that the optimized parameters can be achieved on existing equipment. Standard mandatory requirement boundaries define the mandatory performance indicators required by national or industry standards for coated glass products, which are the basic conditions for product qualification. Mass production environment fluctuation boundaries define the normal fluctuation range of environmental parameters in the coating production workshop, including ambient temperature, ambient humidity, and ambient cleanliness, providing an environmental adaptation basis for parameter optimization.

[0033] In some embodiments, historical operating status data of the coating equipment can be collected simultaneously during the data acquisition process, including equipment operating time, target material wear level, cavity sealing performance, supplementing the operating condition boundary data content, and improving the completeness of the collected data.

[0034] S1.2. Standardization preprocessing of collected data: This step involves cleaning and normalizing the collected raw data to eliminate invalid content, missing content, and dimensional differences, ensuring that the processed data can be directly used for database construction and subsequent algorithm optimization. This includes three steps: outlier removal, missing value completion, and dimensionless normalization.

[0035] Outlier removal is achieved using anomaly filtering rules based on the normal distribution characteristics of the data. Distribution characteristic analysis is performed on each type of raw data, and corresponding distribution intervals are defined. Invalid collected data exceeding these intervals are removed to ensure the validity of the remaining data. Multiple imputation is used to complete missing values, filling in any missing fields encountered during data collection to ensure the completeness of the processed data samples. Linear transformation standardization is employed to achieve dimensionless processing, mapping all collected data to a set numerical interval. This eliminates interference from data with different dimensions in subsequent optimization processes, ensuring that all processed data are on the same numerical scale and can be directly used for subsequent algorithm calculations and optimizations.

[0036] In some embodiments, during the outlier removal process, auxiliary judgment can be made by combining the records of abnormal events in the production process. For corresponding data samples marked with production abnormalities or detection abnormalities, they can be directly judged as outliers and removed, thereby improving the efficiency and accuracy of outlier removal.

[0037] S1.3. Data Classification, Storage, and Database Setup: This step categorizes and stores the preprocessed valid data according to set rules, forming a structured dataset to provide quickly retrieved and reusable data support for subsequent parameter optimization. Data is categorized and stored according to a three-level classification rule: substrate type, coating material system, and application scenario. This rule is a hierarchical classification storage rule with substrate type as the first level, coating material system as the second level, and application scenario as the third level. The first level corresponds to an independent storage partition, the second level to independent sub-partitions within the partition, and the third level to independent data units within the sub-partitions. Specific classification combinations and corresponding storage contents are shown in Table 1. Under the first-level float glass partition, two sub-partitions are set up: single-dielectric film for construction and metal-dielectric composite film for automotive, storing the corresponding combination's process parameters, performance test data, and operating condition boundary data, respectively. Under the first-level tempered glass partition, a metal-dielectric composite film for construction sub-partition is set up, storing the corresponding three types of data. Under the first-level ultra-clear glass partition, two sub-partitions are set up: multilayer laminated film for home appliances and metal-dielectric composite film for construction, storing the corresponding three types of data. Each data unit is assigned a unique retrieval code, which consists of three fixed-length characters: substrate type, coating material system, and application scenario. The corresponding data sample can be quickly retrieved through the code.

[0038] Table 1. Three-level classification storage table for the full-element database of coating. First-level classification Second-level classification Third-level classification Storage data types float glass Single dielectric membrane Construction Process parameters, performance test data, and operating condition boundary data float glass Metal-dielectric composite film automotive Process parameters, performance test data, and operating condition boundary data tempered glass Metal-dielectric composite film Construction Process parameters, performance test data, and operating condition boundary data Ultra-clear glass Multilayer laminated membrane Home appliances Process parameters, performance test data, and operating condition boundary data Ultra-clear glass Metal-dielectric composite film Construction Process parameters, performance test data, and operating condition boundary data Each set of valid data after preprocessing is classified and labeled according to the three-level classification rules, stored in the corresponding storage unit, and a standardized database of all elements of heat-insulating glass coating is built.

[0039] S1.4. Iterative updates of the database: After a single coating production is completed, the process parameters, performance test data, and operating condition data corresponding to this production are standardized and preprocessed, and then synchronized to the standardized database of all elements of heat insulation glass coating. The database is classified and stored according to the three-level classification rules, and the database is iteratively updated. As the production process continues, the number of effective samples in the database continues to increase, which can improve the accuracy and convergence efficiency of subsequent algorithm optimization.

[0040] In some specific implementations, after the initial database setup was completed, 1200 sets of valid historical production data samples were collected. These samples included three types of substrates (float glass, tempered glass, and ultra-clear glass), three types of coating materials (single-dielectric film, metal-dielectric composite film, and multilayer laminated film), and three application scenarios (construction, automotive, and home appliances). Outlier removal was performed on all collected samples, eliminating 32 sets of samples that exceeded the set distribution range. Missing values ​​were filled in for the remaining 1168 sets of samples, completing 12 missing fields. Dimensionless normalization was then performed, mapping all data to a numerical range of 0 to 1. The data was then classified and stored according to a three-level classification rule, with each data unit containing at least 50 samples to ensure sufficient sample size for subsequent algorithm initialization. A 12-bit retrieval code was assigned to each data unit: the first four bits represent the substrate type, the middle four bits represent the coating material system, and the last four bits represent the application scenario. This coding allows for the retrieval of the corresponding data sample within 100ms, improving data reuse efficiency.

[0041] S2. Construction of a multi-objective optimization system: S2.1. Optimize the construction of variable parameter vectors: This step selects the coating process parameters that need to be adjusted and optimized, and constructs the corresponding parameter vector, providing a foundation for subsequent objective function construction and algorithm optimization. From the controllable process parameters stored in the standardized database of all elements of heat-insulating glass coating, controllable coating process parameters are selected as optimization variables, and an optimization variable parameter vector is constructed. Each dimension in the parameter vector corresponds to a selected controllable coating process parameter. The number of dimensions in the parameter vector is consistent with the number of selected optimization variables. The value range of each dimension corresponds to the feasible interval of the corresponding controllable parameter. The feasible interval is determined by the operating condition boundary data stored in the database, ensuring that the values ​​of the optimization variables are always within the achievable range of the equipment.

[0042] S2.2. Construction of the overall objective function for multi-objective optimization: This step constructs mathematical expressions to evaluate the merits of combinations of process parameters, providing a unified computational basis for the algorithm optimization process. Based on three dimensions—film performance, production efficiency, and production energy consumption—corresponding sub-objective functions are constructed, including a performance sub-objective function, a production efficiency sub-objective function, and a production energy consumption sub-objective function. The performance sub-objective function is constructed based on the spectral performance and uniformity indicators of the film layer; the production efficiency sub-objective function is constructed based on the travel speed and number of debugging iterations in the coating production process; and the production energy consumption sub-objective function is constructed based on the energy consumption indicators of various equipment during the coating process.

[0043] In some embodiments, a linear weighting method is used to combine multiple sub-objective functions to construct a multi-objective optimization overall objective function. The linear weighting method is a processing method that assigns a corresponding weight coefficient to each sub-objective function and merges multiple sub-objective functions into a single objective function.

[0044] In some embodiments, the specific calculation process is as follows: First, the original calculation results of each sub-objective function are normalized and mapped to a numerical range of 0 to 1 to eliminate the dimensional differences between different sub-objective functions; then, each normalized sub-objective function result is multiplied by its corresponding weight coefficient; finally, all the multiplied results are summed to obtain the final calculation result of the overall multi-objective optimization objective function. The sum of the weight coefficients corresponding to the linear weighting method is 1. The weight coefficients are adjusted according to the application scenario of the coating product. The sub-objective weight allocation corresponding to different application scenarios is shown in Table 2. For coating products used in construction, the performance target weight is set to 0.5, the production efficiency target weight to 0.3, and the production energy consumption target weight to 0.2, adapting to the balanced requirements of film performance and production efficiency in construction scenarios. For coating products used in automobiles, the performance target weight is increased to 0.6, the production efficiency target weight is set to 0.2, and the production energy consumption target weight remains at 0.2, adapting to the higher performance requirements of automotive scenarios. For coating products used in home appliances, the performance target weight is set to 0.4, the production efficiency target weight is increased to 0.4, and the production energy consumption target weight remains at 0.2, adapting to the higher production efficiency requirements of home appliance scenarios. For coating products used in special applications, the performance target weight is further increased to 0.7, the production efficiency target weight is set to 0.2, and the production energy consumption target weight is set to 0.1, adapting to the core performance requirements of special applications. Based on the weight allocation for the corresponding scenarios, the overall objective function for multi-objective optimization is constructed.

[0045] Table 2 Weight Allocation Table for Sub-Objective Functions in Multi-Objective Optimization Application scenarios Performance-based target weights Production efficiency is divided into target weights Production energy consumption target weight Construction 0.5 0.3 0.2 automotive 0.6 0.2 0.2 Home appliances 0.4 0.4 0.2 Special Purpose 0.7 0.2 0.1 In some embodiments, a sub-objective function related to the weather resistance of the membrane can be added and incorporated into the construction process of the overall objective function of multi-objective optimization. By adjusting the weighting coefficients, the long-term performance of the membrane can be taken into account.

[0046] S2.3. Setting of hard constraints: This step sets inviolable mandatory boundary conditions during parameter optimization to ensure the feasibility and compliance of the optimized process parameter combinations. Combining the equipment rated parameter boundaries and standard mandatory requirement boundaries from the comprehensive standardized database for heat-insulating glass coating, hard constraints are set for the optimization process. The hard constraints corresponding to the equipment rated parameter boundaries limit the value range of each optimization variable, ensuring that the optimized parameters can be achieved on existing coating equipment. The hard constraints corresponding to the standard mandatory requirement boundaries limit the range of basic performance indicators for the coated product, ensuring that the product corresponding to the optimized parameter combination meets the mandatory requirements of the standard.

[0047] S2.4. Setting soft constraints: This step sets the process stability-related boundary conditions that need to be met during parameter optimization to ensure that the optimized process parameter combination has good mass production stability. Combining film performance index data and production process index data from the standardized database of all elements of heat-insulating glass coating, soft constraints are set for the optimization process. These soft constraints include the range of film thickness deviation, the range of surface resistivity non-uniformity, the range of product yield, and limitations on process defects such as film cracking, peeling, and optical distortion. By setting these soft constraints, process parameter combinations with good mass production stability are selected.

[0048] S3. Particle Swarm Optimization Algorithm: S3.1. Population initialization for particle swarm optimization: This step involves constructing an initial particle swarm before the algorithm iteration begins, providing an initial search starting point for the algorithm's optimization. Historical process parameter combinations for the corresponding substrate type and coating material system are retrieved from the standardized database of all elements of heat-insulating glass coating. The particle swarm algorithm's population initialization is then completed based on these retrieved parameter combinations.

[0049] In some embodiments, each particle corresponds to a set of coating process parameters, the particle's position corresponds to the value of the optimization variable parameter vector, the particle's velocity corresponds to the adjustment step size and direction of the optimization variable, and the particle's dimension is consistent with the dimension of the optimization variable parameter vector. Before running the particle swarm optimization algorithm, the basic hyperparameters are configured. Specific hyperparameter settings are shown in Table 3: the population size is set to 50, determining the initial search range for a single iteration; the maximum number of iterations is set to 100, limiting the maximum computational load of the algorithm; the individual learning factor is set to 1.5 to balance the influence of the particle's own historical optimization experience on the iteration process; the social learning factor is set to 1.7 to balance the influence of the entire population's optimization experience on the iteration process; the upper limit of the inertia weight is set to 0.9 to control the algorithm's global search capability in the early stages of iteration; and the lower limit of the inertia weight is set to 0.4 to control the local optimization accuracy in the later stages of iteration. Based on the above hyperparameter configuration, the initialization settings of the particle swarm optimization algorithm are completed, assigning an initial position and initial velocity to each particle in the population.

[0050] Table 3. Basic Hyperparameter Settings for Particle Swarm Optimization Algorithm Hyperparameter name Parameter meaning Parameter values Parameter Function Population size Total number of particles in a single iteration 50 Determine the initial search range of the algorithm Maximum number of iterations Maximum number of iterations of the algorithm 100 Limit the maximum computational cost of the algorithm Individual learning factors Weights learned by a particle from its own optimal solution 1.5 Balancing the influence of individual experience Social learning factors Weights learned by particles from the population optimal solution 1.7 The influence of balancing group experience Inertia weight limit Maximum inertia weight in the early stage of iteration 0.9 Controlling global search capabilities in the early stages of iteration Lower bound of inertia weight Minimum inertia weight in the later stage of iteration 0.4 Controlling local optimization capabilities in the later stages of iteration In some embodiments, during the construction of the initial population, some random particles can be generated within the neighborhood of the retrieved historical process parameter combinations and added to the initial population to expand the search range of the initial population and avoid the algorithm getting stuck in local optima.

[0051] S3.2. Calculation of particle fitness value: This step calculates the fitness value of each particle in the population, providing a basis for subsequent optimal solution updates. The parameter vector corresponding to each particle is substituted into the overall objective function of the multi-objective optimization, and the fitness value of each particle is calculated in conjunction with the set constraints. For particles that satisfy all constraints, their fitness values ​​are directly adopted from the calculation result of the overall objective function; for particles that do not satisfy the constraints, their fitness values ​​are adjusted through a penalty mechanism. The penalty mechanism is to add a penalty term positively correlated with the degree of deviation from the constraints to the overall objective function calculation result for particles that do not satisfy the constraints, making their fitness values ​​worse than those of particles that satisfy the constraints, and gradually eliminating them during the iteration process. In each iteration, the fitness value of all particles in the population is calculated.

[0052] S3.3. Update of Individual Optimal Solution and Global Optimal Solution: This step updates the individual optimal solution and the global optimal solution during the iteration process, driving the algorithm towards a better parameter combination. The individual optimal solution refers to the position with the best fitness value that a single particle has historically encountered during the iteration process, corresponding to the optimal combination of process parameters found by that particle. The global optimal solution refers to the position with the best fitness value that the entire particle population has historically encountered during the iteration process, corresponding to the optimal combination of process parameters found by the entire population. By comparing the current fitness value of each particle with its historical best fitness value, if the current fitness value is better, the particle's current position is updated to the new individual optimal solution. The individual optimal solutions of all particles in the population are traversed, and the individual optimal solution with the best fitness value is selected and updated as the new global optimal solution. The updates of the individual optimal solution and the global optimal solution are completed in each iteration.

[0053] S3.4. Particle velocity and position update and population diversity maintenance: This step iteratively updates the particle velocity and position while maintaining population diversity to prevent the algorithm from getting trapped in local optima. Based on an adaptive inertia weight adjustment mechanism, the velocity and position of each particle are updated. This mechanism dynamically adjusts the inertia weight in the particle swarm velocity update formula as the iteration progresses. Specifically, the inertia weight is adjusted as follows: the initial value is a set upper limit weight value, which is gradually reduced to the lower limit weight value in a linear manner as the number of iterations increases. Simultaneously, for a single particle, if its current fitness value is better than the population average fitness value, its inertia weight is reduced by an adjustment coefficient less than 1 based on the current iteration's inertia weight value, preserving its current favorable position; if its current fitness value is worse than the population average fitness value, its inertia weight is increased by an adjustment coefficient greater than 1, propelling it towards a better position.

[0054] The particle velocity update formula consists of three parts: the product of the particle's current velocity and inertia weight, an individual learning term, and a social learning term. These three parts are added together to obtain the updated particle velocity. The updated velocity is limited to a set velocity range to prevent the particle from exceeding the parameter search range due to excessive velocity. The particle position update formula is the sum of the particle's current position and the updated velocity to obtain the updated position. The updated position is limited to the feasible range of the optimization variables to ensure that the parameter combination corresponding to the particle position meets the hard constraints of the device's rated parameters.

[0055] Population diversity is characterized by the variance of the fitness values ​​of all particles in the population. The smaller the variance of the fitness values, the lower the population diversity. Population diversity is monitored in real time. When the population diversity falls below a set threshold, it is determined that the population is trapped in a local optimum. A Gaussian mutation operation is then performed on a randomly selected subset of particles. The Gaussian mutation operation involves adjusting the values ​​of some dimensions of the particles and superimposing a random number that conforms to a Gaussian distribution. This adjusts the particle positions. The mean of the Gaussian distribution is 0, and the variance is dynamically adjusted based on the number of iterations. In the early stages of iteration, the variance is larger, expanding the mutation range; in the later stages, the variance is smaller, ensuring mutation accuracy. This mutation operation reactivates the population diversity, propelling the algorithm out of local optima.

[0056] S3.5. Iteration Termination Judgment and Optimal Parameter Output: This step determines the iteration termination condition. The iteration terminates when the termination condition is met, and the optimal combination of process parameters is output. There are two types of iteration termination conditions: the first is reaching the set maximum number of iterations; the second is that the fitness value of the global optimal solution changes less than a set convergence accuracy threshold over multiple consecutive iterations. Meeting either condition terminates the iterative optimization process, and the coating process parameter combination corresponding to the global optimal solution obtained during the iteration is output. The iteration termination condition is set before the iteration begins, and it is checked after each iteration to ensure that the algorithm terminates the iteration promptly after finding the optimal parameter combination that meets the accuracy requirements.

[0057] In some specific implementations, after setting the algorithm hyperparameters, the population is initialized based on 80 sets of historical high-quality parameter combinations retrieved from the database. In the initial population, 40% of the particles directly use the positions corresponding to the historical high-quality parameter combinations, while 60% of the particles are randomly generated within the neighborhood of the high-quality parameter combinations. The neighborhood range is set to ±15% of the high-quality parameter values, ensuring both the effectiveness of the initial population and preserving a sufficient search range. During the iteration process, population diversity is checked every 5 iterations. When the variance of the population fitness value is lower than a set threshold of 0.001, Gaussian mutation is performed on 10% of the particles in the population. When the iteration reaches the 32nd generation, if the change in the fitness value of the global optimum over 10 consecutive generations is less than the convergence accuracy threshold of 1e-6, the iteration termination condition is met, the iteration is terminated, and the corresponding coating process parameter combination is output. This effectively avoids premature convergence and ensures that the algorithm can find high-quality parameter combinations globally.

[0058] S4. Coating parameter verification and correction: S4.1. Optical simulation verification of coating parameters: This step uses optical simulation to initially verify the algorithm's output parameters, selecting parameter combinations that meet the constraints, avoiding invalid experiments, and reducing experimental losses. The output coating process parameter combinations are imported into optical thin film simulation software to perform simulation calculations of the film's optical performance, obtaining optical performance indicators such as the film's spectral curve, visible light transmittance, and infrared blocking rate, verifying whether the coating process parameter combinations meet the set constraints. For coating process parameter combinations that meet the constraints in simulation verification, they proceed to the subsequent experimental verification stage; for those that do not meet the constraints, the weight coefficients of the multi-objective optimization overall objective function and the boundary range of the constraints are adjusted. The adjusted multi-objective optimization overall objective function and constraints are then substituted into the particle swarm optimization algorithm, and the complete iterative optimization process is re-executed to output new coating process parameter combinations. The simulation verification process is then repeated until the simulation verification meets the constraints.

[0059] S4.2. Laboratory Small-Scale Experiment Verification of Coating Process: This step verifies the actual performance of the parameter combination through laboratory coating experiments, connecting algorithm optimization with mass production implementation. For coating process parameter combinations that meet the constraints verified in simulation, small-scale laboratory coating tests are performed. After completing the coating tests according to the parameter combinations, corresponding testing equipment is used to conduct actual tests on the film performance and production indicators. The test items include the film's optical properties, mechanical properties, weather resistance, surface resistivity uniformity, as well as indicators such as energy consumption and debugging time during the production process, obtaining the actual test results corresponding to this parameter combination.

[0060] S4.3. Coating parameter deviation correction and iterative optimization: This step compares the pilot test results with the constraints, corrects and re-optimizes the parameters to ensure that the final output parameter combination meets the optimization objectives in actual production. If deviations exist, the pilot test data is standardized and preprocessed before being synchronized to the standardized database for all elements of the heat-insulating glass coating. The iterative optimization process of the particle swarm optimization algorithm is then re-executed, outputting the corrected coating process parameter combination. Simulation verification and pilot test verification are then repeated until the pilot test results meet the requirements of the constraints.

[0061] S4.4. Final coating process parameter output: For coating process parameter combinations that meet the constraints in small-scale trials, the output is the final, feasible coating process parameter combination. This parameter combination, after algorithm optimization, simulation verification, small-scale trials, and deviation correction, can be directly used for mass production of coated products.

[0062] S4.5. Dynamic adjustment of parameters and database update during mass production: This step involves dynamically adjusting parameters during mass production to compensate for performance deviations caused by fluctuations in operating conditions, while simultaneously updating the database. The final, feasible coating process parameter combination is imported into the mass production coating line. Real-time data on operating condition fluctuations during mass production is collected, and micro-iterative optimization is performed on the coating process parameter combination based on this data. Micro-iterative optimization maintains the core framework of the original coating process parameter combination, only iteratively optimizing parameters affected by operating condition fluctuations within a small scope, thus completing the dynamic parameter adjustment during mass production. Simultaneously, the process parameters, performance testing data, and operating condition data from the mass production process are standardized and preprocessed, then synchronized to the standardized database of all elements for heat-insulating glass coating, ensuring continuous iterative updates to the database and constantly enriching its sample size.

[0063] In some embodiments, during mass production, the coating process parameters can be periodically optimized using micro-iterations to address the target material loss during continuous production, thereby compensating for the changes in coating effect caused by target material loss and improving the consistency of film performance during continuous production.

[0064] This solution effectively reduces the influence of human experience on coating process parameter design through end-to-end parameter optimization, improves the standardization of coating process parameter design, and enhances the consistency and stability of coating film performance. By establishing and iteratively updating a standardized database of all elements, effective reuse of relevant data throughout the coating production process is achieved, shortening the debugging cycle of process parameters, reducing raw material losses during production changeovers, and improving the operational efficiency of the coating production line to a certain extent. The construction of a multi-objective optimization system addresses the optimization needs of multiple dimensions, including film performance, production efficiency, and production energy consumption, avoiding performance deficiencies in other dimensions caused by single-dimensional optimization, and achieving synergistic optimization of multiple coating process parameters. Iterative optimization using particle swarm optimization addresses the challenge of optimizing coupled multi-parameter coating processes, obtaining better-suited process parameter combinations and improving the overall performance of coated products. A closed-loop process of simulation verification, small-scale experimental verification, and deviation correction ensures the practical feasibility of the optimized process parameter combinations, reducing the frequency of process adjustments during mass production. This solution is adaptable to different types of coating processes, substrates and coating material systems, and has good versatility and scenario adaptability, which can improve the overall efficiency of coating production to a certain extent.

[0065] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A particle swarm optimization design method for thermal insulation glass coating parameters, characterized in that, Includes the following steps: S1. Collect basic material characteristic data, process parameter data, performance test data and working condition boundary data related to heat insulation glass coating, preprocess all collected data, and build a standardized database of all elements of heat insulation glass coating according to the set classification rules. S2. Select controllable parameters of the coating process as optimization variables, construct a multi-objective optimization objective function, and combine the working condition boundary data in the standardized database of all elements of heat insulation glass coating to set the constraints of the optimization process; S3. Based on the standardized database of all elements of heat insulation glass coating, the population initialization of the particle swarm algorithm is completed. Combined with the multi-objective optimization total objective function and constraint conditions, iterative optimization is performed to update the individual optimal solution and the global optimal solution of the particle swarm. After the iteration terminates, the coating process parameter combination corresponding to the global optimal solution is output. S4. Perform simulation and experimental verification on the output coating process parameter combination, correct the parameters based on the verification results, and output the final feasible coating process parameter combination.

2. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, Step S1 includes: S1.

1. Collect basic material characteristic data, process parameter data, performance test data, and operating condition boundary data; basic material characteristic data includes substrate parameters, coating material parameters, and auxiliary material parameters; process parameter data includes controllable process parameters and auxiliary process parameters; performance test data includes film performance index data and production process index data; operating condition boundary data includes equipment rated parameter boundaries, standard mandatory requirement boundaries, and mass production environment fluctuation boundaries. S1.

2. Perform outlier removal, missing value completion, and dimensionless normalization on all collected data to obtain standardized and valid data; S1.

3. According to the three-level classification rule of substrate type, coating material system and application scenario, the effective data is classified and stored. The three-level classification rule is a hierarchical classification and storage rule with substrate type as the first level, coating material system as the second level and application scenario as the third level, and a standardized database of all elements of heat insulation glass coating is built. S1.

4. After completing a single coating production, synchronize the actual production process parameters, performance test data and operating condition data to the standardized database of all elements of heat insulation glass coating, and complete the iterative update of the standardized database of all elements of heat insulation glass coating.

3. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, Step S2 includes: S2.

1. Select controllable parameters of the coating process as optimization variables from the controllable process parameters stored in the standardized database of all elements of heat insulation glass coating, and construct an optimization variable parameter vector; S2.

2. Based on the three dimensions of membrane performance, production efficiency and production energy consumption, corresponding sub-objective functions are constructed respectively. A linear weighting method is used to combine the sub-objective functions. The linear weighting method is a processing method that assigns a corresponding weight coefficient to each sub-objective function and merges multiple sub-objective functions into a single objective function, thus constructing a multi-objective optimization overall objective function. S2.

3. Based on the equipment rated parameter boundaries and standard mandatory requirement boundaries in the full-element standardized database of heat insulation glass coating, set hard constraints for the optimization process; S2.

4. Based on the film performance index data and production process index data in the standardized database of all elements of heat insulation glass coating, set soft constraints for the optimization process.

4. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, Step S3 includes: S3.

1. Retrieve the historical process parameter combinations of the corresponding substrate type and coating material system from the standardized database of all elements of heat insulation glass coating, and complete the population initialization of particle swarm algorithm based on the retrieved parameter combinations, and assign initial position and initial velocity to each particle in the population. S3.

2. Substitute the parameter vector corresponding to each particle into the overall objective function of the multi-objective optimization, and calculate the fitness value of each particle in combination with the constraints; S3.

3. Compare the current fitness value of each particle with the historical best fitness value to update the individual optimal solution; traverse the individual optimal solutions of all particles in the population to update the global optimal solution; S3.

4. Based on the adaptive inertia weight adjustment mechanism, update the velocity and position of each particle; the adaptive inertia weight adjustment mechanism is to dynamically adjust the inertia weight in the particle swarm velocity update formula according to the number of iterations during the iteration process. The upper limit weight value is set in the early stage of the iteration, and the lower limit weight value is set in the later stage of the iteration; detect the population diversity, which is characterized by the variance of the fitness values ​​of all particles in the population. When the population diversity is lower than the set threshold, perform mutation operation on some particles; S3.

5. Determine if the iteration termination condition is met. If the termination condition is met, terminate the iteration and output the coating process parameter combination corresponding to the global optimal solution. If the termination condition is not met, return to the fitness value calculation step and continue the iteration.

5. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, Step S4 includes: S4.

1. Import the output coating process parameter combination into the optical thin film simulation software, complete the simulation calculation of the film optical performance, and verify whether the coating process parameter combination meets the constraints. S4.

2. For the combination of coating process parameters that meets the constraints verified by simulation, perform laboratory coating pilot tests to complete the actual testing of film performance and production indicators; S4.

3. Compare the small-scale test results with the constraints. If there is a deviation, synchronize the small-scale test data to the standardized database of all elements of heat-insulating glass coating and re-execute the iterative optimization. S4.

4. For small-scale experiments that verify the coating process parameter combinations that meet the constraints, output the final, feasible coating process parameter combinations.

6. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, In step S1, outlier values ​​in the collected data are removed using an anomaly filtering rule based on the normal distribution characteristics of the data, eliminating invalid collected data that exceeds the set distribution range; missing values ​​in the collected data are filled using a multiple interpolation method, filling in missing data in the fields that occur during the collection process; and dimensionless processing of the collected data is completed using a linear transformation standardization method, uniformly mapping all collected data to a set numerical range to eliminate the interference of data with different dimensions on the subsequent optimization process.

7. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, In step S2, the sub-objective functions include a performance sub-objective function, a production efficiency sub-objective function, and a production energy consumption sub-objective function; the sum of the weight coefficients corresponding to the linear weighting method is 1, and the weight coefficients are adjusted according to the application scenario of the coated product; each sub-objective function corresponds to a set calculation logic, and the sub-objective functions are merged into a single multi-objective optimization overall objective function through the linear weighting method.

8. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, In step S3, the iteration termination conditions include two types: the first type is that the number of iterations reaches the set maximum number of iterations, and the second type is that the fitness value of the global optimal solution changes less than the set convergence accuracy threshold for multiple consecutive generations. If either of the two types of iteration termination conditions is met, the iterative optimization process terminates, and the combination of coating process parameters corresponding to the global optimal solution obtained during the iteration process is output.

9. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, In step S4, the simulation verifies the coating process parameter combination that does not meet the constraints. The weight coefficients corresponding to the multi-objective optimization total objective function and the boundary range of the constraints are adjusted. The adjusted multi-objective optimization total objective function and constraints are substituted into the particle swarm optimization algorithm, and the complete iterative optimization process is re-executed to output a new coating process parameter combination. The simulation verification process is then re-executed.

10. The particle swarm optimization design method for thermal insulation glass coating parameters according to claim 1, characterized in that, After step S4 is completed, the final feasible coating process parameter combination is imported into the mass production coating production line. The operating condition fluctuation data during the mass production process is collected in real time. Based on the operating condition fluctuation data, micro-iterative optimization is performed on the coating process parameter combination. Micro-iterative optimization is an optimization process that keeps the main framework of the original coating process parameter combination unchanged and only performs small-scale iterative optimization on the parameter dimensions affected by operating condition fluctuations, thus completing the dynamic adjustment of parameters in the mass production process. The process parameters, performance test data, and operating condition data during mass production are synchronized to the standardized database of all elements of heat-insulating glass coating, and the database is continuously iterated and updated.

Images