Component intelligent optimization method and system for anti-condensation insulation coating

Abstract

The application discloses a kind of component intelligent optimization method and system of anti-condensation insulating paint, it is related to data processing technical field, comprising: establishing the component database of anti-condensation insulating paint, determine component initial range and establish component-performance prediction model;With the target performance parameter of application scene analysis as target, the component initial range is optimized by component-performance prediction model, and the predicted theoretical component is obtained;According to the target performance parameter, the reliability deviation peak value analysis is carried out to the component of predicted theoretical component and initial component, and the test verification component is positioned;Test verification is carried out based on test verification component, and the final optimized target paint component is determined.The application solves the technical problems that anti-condensation insulating paint formula optimization in prior art relies on artificial experience and is difficult to realize multi-performance index collaborative optimization, achieves the technical effect of intelligent collaborative optimization of multi-objective performance and improves the efficiency and accuracy of formula optimization.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a method and system for intelligent optimization of the components of an anti-condensation insulating coating. Background Technology

[0002] In the research and development of anti-condensation insulating coatings, formulation design typically relies on technicians adjusting component ratios and conducting repeated trials based on experience. Complex coupling relationships exist between different components, and improving anti-condensation performance often affects electrical insulation performance, durability, or application performance, making it difficult to simultaneously achieve multiple performance indicators. Due to a lack of systematic data support and quantitative analysis methods, the formulation optimization process is time-consuming, costly, and lacks stability, making it difficult to achieve synergistic improvement and precise control of multiple performance objectives. Summary of the Invention

[0003] This application provides a method and system for intelligent optimization of the components of an anti-condensation insulating coating, which addresses the technical problem that the optimization of anti-condensation insulating coating formulations in the prior art relies on manual experience and is difficult to achieve synergistic optimization of multiple performance indicators.

[0004] In view of the above problems, this application provides a method and system for intelligent optimization of the components of anti-condensation insulating coating.

[0005] The first aspect of this application provides a method for intelligently optimizing the composition of an anti-condensation insulating coating, the method comprising:

[0006] A component database for anti-condensation insulating coatings is established by collecting physicochemical parameter data of the base resin, modified filler, functional additives, and curing system, as well as historical formulation-performance correlation data. Based on the component database, an initial range of components is determined and a component-performance prediction model is established. Taking the target performance parameters analyzed in the application scenario as the target, the initial range of components is optimized through the component-performance prediction model to obtain the predicted theoretical components. Based on the target performance parameters and the components of the predicted theoretical components and the initial components, a reliability deviation peak analysis is performed to locate the test and verification components. Based on the test and verification components, experimental testing and verification are carried out to determine the final optimized target coating components.

[0007] A second aspect of this application provides a component intelligent optimization system for anti-condensation insulating coatings, the system comprising:

[0008] The system comprises the following modules: a database establishment module for collecting physicochemical parameter data of the base resin, modified filler, functional additives, and curing system, as well as historical formulation-performance correlation data, to establish a component database for anti-condensation insulating coatings; a model establishment module for determining the initial range of components and establishing a component-performance prediction model based on the component database; an optimization module for optimizing the initial range of components using the component-performance prediction model, with the target performance parameters analyzed in the application scenario as the objective, to obtain the predicted theoretical components; an analysis module for performing reliability deviation peak analysis based on the target performance parameters and the predicted theoretical components and initial components, to locate the test and verification components; and a test and verification module for conducting experimental tests and verifications based on the test and verification components to determine the final optimized target coating components.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application collects physicochemical parameter data of the base resin, modified filler, functional additives, and curing system, as well as historical formulation-performance correlation data, to establish a component database for anti-condensation insulating coatings. Based on the component database, an initial range of components is determined and a component-performance prediction model is established. Using the target performance parameters analyzed in the application scenario as the objective, the initial range of components is optimized through the component-performance prediction model to obtain predicted theoretical components. Based on the target performance parameters and the components of the predicted theoretical components and the initial components, a reliability deviation peak analysis is performed to locate the test verification components. Based on the test verification components, experimental testing and verification are conducted to determine the final optimized target coating components. This invention solves the technical problem in the prior art where anti-condensation insulating coating formulation optimization relies on manual experience and is difficult to achieve synergistic optimization of multiple performance indicators. By establishing a component database and constructing a component-performance prediction model for model-driven optimization analysis, it achieves intelligent synergistic optimization of multiple target performances and improves the efficiency and accuracy of formulation optimization. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic diagram of a method for intelligent optimization of the components of an anti-condensation insulating coating provided in this application embodiment;

[0013] Figure 2 This is a schematic diagram of the intelligent optimization system structure for the components of an anti-condensation insulating coating provided in an embodiment of this application.

[0014] Figure labeling: Database creation module 11, Model creation module 12, Optimization module 13, Analysis module 14, Testing and verification module 15. Detailed Implementation

[0015] This application provides a method and system for intelligent optimization of the components of anti-condensation insulating coatings. It addresses the technical problem that the optimization of anti-condensation insulating coating formulations in the prior art relies on manual experience and is difficult to achieve synergistic optimization of multiple performance indicators. By establishing a component database and constructing a component-performance prediction model for model-driven optimization analysis, it achieves the technical effect of intelligent synergistic optimization of multiple target performances and improves the efficiency and accuracy of formulation optimization.

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0018] Example 1, as Figure 1 As shown, this application provides a method for intelligent optimization of the components of an anti-condensation insulating coating, the method comprising:

[0019] Step S100: Collect physicochemical parameter data of matrix resin, modified filler, functional additives, and curing system, as well as historical formula-performance correlation data, to establish a component database for anti-condensation insulating coatings.

[0020] In this embodiment, when collecting physicochemical parameter data of the matrix resin, modified filler, functional additives, and curing system, each component is first numbered and dried, dehydrated, and batch-identified. Then, existing standard testing methods are used to obtain the physicochemical parameter data. For the matrix resin, molecular weight is obtained by gel permeation chromatography (GPC), glass transition temperature (Tg) is obtained by differential scanning calorimetry (DSC), and viscosity data is obtained by rotational rheometer. For the modified filler, particle size distribution (D50) is obtained by laser particle size analyzer or dynamic light scattering (DLS), specific surface area is obtained by BET nitrogen adsorption method, and surface functional groups are analyzed by Fourier transform infrared spectroscopy (FTIR) or X-ray photoelectron spectroscopy (XPS). For the functional additives, surface tension and rheological properties are obtained by surface tension meter and rheometer. For the curing system, curing exothermic peak temperature and reaction kinetic parameters are obtained by DSC isothermal scanning. The above test data are standardized by unit unification, outlier removal, and other standardization processes to form a structured physicochemical parameter data record.

[0021] When collecting historical formula-performance correlation data, formula data extraction, performance test data matching, and consistency verification are performed. Formula data extraction includes extracting the content ratios of each component and key process parameters from historical formulas and binding them to formula and batch numbers to obtain a historical formula data table. Performance test data matching involves summarizing performance test results such as volume resistivity, contact angle, elongation at break, and pressure resistance tests from the same batch and binding them to test numbers, then matching each result with the formula number to obtain a historical performance data table. Consistency verification involves verifying the repeatability of different batches of the same formula, marking abnormal batches, retaining traceable records, and obtaining cleaned historical formula-performance correlation data.

[0022] Finally, a component database for anti-condensation insulating coatings was established. This process began by first organizing historical formulation-performance correlation data based on existing research and experimental datasets, comparing the content of each component with the corresponding coating performance, and analyzing the degree of influence of each component on coating performance. Simultaneously, combined with data from different production batches, the performance data of the same component in actual production was statistically analyzed, calculating the standard deviation and coefficient of variation (CV) to analyze the performance fluctuations of the component in actual production. Based on this, according to the degree of influence of each component on coating performance and its performance fluctuations in actual production, the components were classified into stable key components with significant performance impact and low fluctuations, suspected key components with significant performance impact and high fluctuations, stable auxiliary components with small performance impact and low fluctuations, and suspected auxiliary components with small performance impact and high fluctuations. Finally, according to the influence relationship between each component type and performance parameters, the component types were mapped and correlated with target performance parameters, and the relevant data were integrated to establish a component database for anti-condensation insulating coatings.

[0023] Furthermore, the method provided in the application embodiments, in establishing a component database of anti-condensation insulating coatings, also includes:

[0024] Based on existing research and experimental datasets, the influence of each component on coating performance and the performance fluctuation of the components in actual production are analyzed. The components are classified into types, including stable key components with large performance impact and small fluctuation, suspected key components with large performance impact and large fluctuation, stable auxiliary components with small performance impact and small fluctuation, and suspected auxiliary components with small performance impact and large fluctuation. According to the influence relationship between each component type and performance parameters, the component types are mapped and associated with target performance parameters to establish a component database for anti-condensation insulating coatings.

[0025] In this embodiment of the application, based on the existing research and experimental dataset, the historical formulation-performance correlation data is first uniformly organized, and the component content of the matrix resin, modified filler, functional additive, and curing system is matched with the corresponding performance parameters one by one to form a complete set of data records. On this basis, the performance parameter changes of each component under different content ranges are statistically analyzed, the performance parameter change range caused by the change of component content is calculated, and the change range of all sample ranges is averaged and summarized to obtain the numerical results of the influence of each component under each performance parameter dimension.

[0026] Subsequently, based on the production batch records in the same existing research and experimental dataset, statistical analysis was performed on the key physicochemical parameters and performance parameters of the same component under different batch conditions. The mean and standard deviation were calculated, and the coefficient of variation (CV) was calculated to reflect the fluctuation range. The CV of each component in all batches was summarized to obtain the numerical results of the performance fluctuation of each component in actual production.

[0027] After obtaining the numerical results of the degree of influence and the performance fluctuation, the degree of influence and coefficient of variation (CV) of all components are sorted, and the quartile thresholds are calculated. Components with the degree of influence in the upper quartile and the CV in the lower quartile are classified as stable critical components, components with the degree of influence in the upper quartile and the CV in the upper quartile are classified as suspected critical components, components with the degree of influence in the lower quartile and the CV in the lower quartile are classified as stable auxiliary components, and components with the degree of influence in the lower quartile and the CV in the upper quartile are classified as suspected auxiliary components. This results in a component type classification result that includes a component identifier, a numerical value of the degree of influence, a numerical value of the CV, and a component type field.

[0028] Finally, based on the influence relationship between component type and performance parameters, the ranking results of the influence degree of each component under different performance parameter dimensions are jointly labeled with the component type field. The mapping relationship between component type and target performance parameter is established respectively. The existing research experimental dataset, influence degree numerical results, performance fluctuation numerical results, component type classification results and component type-target performance parameter mapping relationship are integrated to form a structured component database of anti-condensation insulating coating.

[0029] Step S200: Based on the component database, determine the initial range of components and establish a component-performance prediction model.

[0030] In this embodiment, when determining the initial range of components based on the component database, the statistical distribution of the content of each component in the existing formulation stored in the component database is first obtained. The minimum, maximum and concentrated distribution ranges of the matrix resin, modified filler, functional additives and curing system are extracted respectively. Combined with physicochemical constraints and expert experience, the ranges that do not meet the reaction stoichiometric relationship and process feasibility conditions are eliminated to determine the feasible range of each component. Then, within the feasible range of each component, the basic performance of the anti-condensation insulating coating is used as the benchmark to screen the fixed values ​​or ranges of each component to obtain the fixed values ​​or ranges of each component. Finally, the fixed values ​​or ranges of each component are combined and integrated to form the initial range of components.

[0031] When establishing the component-performance prediction model, historical formulation data and historical formulation-performance correlation data in the component database are used as training samples. The component content is used as the input feature and the target performance parameter is used as the output label. The initial prediction model is trained using machine learning algorithms. Then, the prediction accuracy of the initial prediction model is evaluated using cross-validation. When the prediction accuracy does not meet the preset standard, sparse data regions are identified and experimental points are added. The supplemented experimental data is incorporated into the training samples to retrain the model until the prediction accuracy meets the preset standard, thus obtaining the final component-performance prediction model.

[0032] Furthermore, in the method provided in the application embodiments, determining the initial range of components based on the component database further includes:

[0033] Obtain the statistical distribution of the content of each component in the existing formulas stored in the component database, and determine the feasible range of each component by combining physicochemical constraints and expert experience; based on the feasible range of each component, screen the fixed values ​​or ranges of the performance of each component with the basic performance of the anti-condensation insulating coating as the benchmark, and obtain the fixed values ​​or ranges of each component; determine the initial range of the component according to the fixed values ​​or ranges of each component.

[0034] In this embodiment, to obtain the statistical distribution of the content of each component in existing formulations stored in the component database, historical formulation-performance correlation data is first extracted from the component database. The actual usage of matrix resin, modified filler, functional additives, and curing system in all historical formulations is listed one by one and sorted according to the numerical value. The minimum usage, maximum usage, and commonly used value range in the middle concentration area of ​​each component are determined, thereby obtaining the statistical distribution range of each component in actual application. Subsequently, the statistical distribution range is checked item by item in combination with physicochemical constraints. Physicochemical constraints include the reaction matching relationship between matrix resin and curing system, the effect of filler addition on system fluidity and dispersion uniformity, the system's ability to complete the curing reaction at 40~50℃, and specific indicators such as volume resistivity, contact angle, and elongation at break after film formation meeting basic performance requirements. At the same time, combined with the stable ratio experience formed by long-term production, value ranges that do not meet the reaction matching relationship or cannot meet the basic performance requirements are eliminated. Finally, the feasible range of each component under the reaction conditions, processing conditions, and performance requirements is determined.

[0035] After determining the feasible range for each component, the performance fixed values ​​or ranges of each component are screened based on the basic performance of the anti-condensation insulating coating. The basic performance refers to the formulation level that simultaneously meets the requirements of volume resistivity not less than 1×10¹³Ω·cm, contact angle greater than 120°, and elongation at break not less than 9% in historical formulation-performance correlation data. During the screening process, for stable key components, the actual usage amount in the basic formulation is directly used as the fixed value. If there are fluctuations, the usage amount is limited to a very narrow range within its historical statistical fluctuation range. For suspected key components, a complete segment covering the main historical usage range is selected within its feasible range to retain adjustment space. For stable auxiliary components, the usage amount in the basic formulation is used as the fixed value. For suspected auxiliary components, a narrow segment limited by the historical fluctuation range is selected within the feasible range centered on the historical average usage amount to obtain the fixed values ​​or ranges for each component.

[0036] Finally, based on the fixed values ​​or ranges of each component, the value boundaries of the matrix resin, modified filler, functional additives, and curing system are unified and integrated to form candidate formulation combinations. Subsequently, each candidate formulation combination is checked, including verifying whether the metering match between the matrix resin and the curing system meets the reaction ratio requirements, verifying whether the amount of modified filler added is within the range of uniform dispersion, verifying whether the curing reaction can be completed under the given stirring speed, reaction temperature, and reaction time conditions, and confirming whether the volume resistivity, contact angle, and elongation at break of the combination are within the required range based on historical formulation-performance correlation data. After all the above verification conditions are met, the initial range of components containing the clear value boundaries and fixed values ​​of each component is finally formed.

[0037] Furthermore, the method provided in the application embodiments, in establishing the component-performance prediction model, further includes:

[0038] Using historical formula data and historical formula-performance correlation data from the component database as training samples, with component content as input features and target performance parameters as output labels, an initial prediction model is trained using a machine learning algorithm. The prediction accuracy of the initial prediction model is evaluated using cross-validation. When the prediction accuracy does not meet the preset standard, sparse data regions are analyzed, experimental points are added to the sparse data regions, and the training samples are expanded using the supplemented experimental data. The model is then retrained until the prediction accuracy meets the preset standard, thus obtaining the component-performance prediction model.

[0039] In this embodiment, when using historical formulation data and historical formulation-performance correlation data from the component database as training samples, the matrix resin content, modified filler content, functional additive content, curing system content, and corresponding target performance parameters are first extracted from the component database according to the formulation number. Each formulation is organized into a sample record, where the component content is used as the input feature and the target performance parameter is used as the output label. Subsequently, all samples are preprocessed uniformly, including deleting completely duplicate formulation records, deleting records with missing target performance parameters, and removing data that exceed three times the standard deviation above or below the historical value range of the component. The component content is also standardized by subtracting the sample mean from each component content and then dividing by the standard deviation to eliminate dimensional differences. After preprocessing, all samples are divided into training and validation sets according to a fixed ratio, for example, 80% as the training set and 20% as the validation set, to obtain standardized training samples for model training and evaluation.

[0040] When training the initial prediction model, the multiple linear regression algorithm is selected as the machine learning algorithm. The component content in the training set is used as the input matrix, and the target performance parameters are used as the output vector. The regression coefficients are solved by the least squares method to minimize the sum of the squared errors between the predicted values ​​and the actual target performance parameters. Thus, a mathematical model that can map the component content to the predicted values ​​of the target performance parameters is obtained. This mathematical model is the initial prediction model.

[0041] To evaluate the prediction accuracy of the initial prediction model, a 5-fold cross-validation method was used. The training samples were divided into five subsets on average. In each round, one subset was selected as the validation set, and the remaining four subsets were used as the training set for model refitting. This process was repeated for five rounds. In each round, the model predictions were compared with the actual target performance parameters, and the coefficient of determination R was calculated. 2 As a metric for prediction accuracy, the R² values ​​obtained from 5 rounds of validation were recorded, and the R² values ​​from 5 rounds were taken. 2 The average of the numerical values ​​is used as the model's prediction accuracy; when the prediction accuracy is lower than a preset standard, such as R0... 2 If the value is less than 0.90, the initial prediction model is deemed not to have met the preset standard.

[0042] When the prediction accuracy does not meet the preset standard, a systematic analysis of the component content space is performed. The value range of each component is divided into a fixed number of segments at equal intervals, for example, each component is divided into 5 consecutive segments. Different component segments are then combined, and the number of samples in each segment combination is counted. When the number of samples in a combination is less than 3, the component content interval where the combination is located is determined as a data sparse region. Subsequently, specific component content combinations are selected in this data sparse region according to the segment boundary and the segment center, corresponding formulation samples are prepared, and target performance parameters are tested to form new historical formulation-performance correlation data. The supplementary collected experimental data is added to the original training samples, and the data preprocessing, multiple linear regression training, and 5-fold cross-validation evaluation steps are repeated. The above process is continuously iterated until the prediction accuracy reaches or exceeds the preset standard, and finally a component-performance prediction model that meets the preset standard is obtained.

[0043] Step S300: Taking the target performance parameters analyzed in the application scenario as the objective, optimize the initial range of the components through the component-performance prediction model to obtain the predicted theoretical components.

[0044] In this embodiment of the application, when the target performance parameter is analyzed based on the application scenario, the anti-condensation insulation requirements are first analyzed according to the application environment to determine the required performance of the scenario. Then, the basic performance of the anti-condensation insulating coating is used as a reference to conduct a demand difference analysis to clarify the specific numerical requirements of the target performance parameter. On this basis, the target performance parameter is used as an input condition. Combined with the determined initial range of components, the performance of different component content combinations is predicted and compared and screened through the component-performance prediction model. The initial range of components is optimized, and finally, the predicted theoretical components that meet the requirements of the target performance parameter are obtained.

[0045] Furthermore, in the method provided in the application embodiments, taking the target performance parameters analyzed in the application scenario as the objective, and optimizing the initial range of the components through the component-performance prediction model to obtain the predicted theoretical components, the method further includes:

[0046] Based on the application scenario, the anti-condensation insulation requirements are analyzed to determine the required performance for the scenario. Starting from the basic performance of the anti-condensation insulating coating, the required performance for the scenario is analyzed to determine the target performance parameters. Using the target performance parameters as input, the component performance response is analyzed through the component-performance prediction model to obtain the predicted theoretical components.

[0047] In this embodiment, the anti-condensation insulation requirements are analyzed based on the application scenario. When determining the required performance for the scenario, the environmental and operating condition boundaries of the application scenario are first determined. The relative humidity range, temperature range, conditions for condensation formation on the equipment surface, expected service life, construction method, and acceptable construction time window are recorded sequentially. Then, the above boundary conditions are converted into performance parameter requirements one by one to form a set of required performance for the scenario. Specifically, the anti-condensation requirement is converted into a contact angle requirement, the insulation requirement is converted into a volume resistivity requirement, and the durability requirement is converted into an elongation at break requirement. Each performance parameter is written as a clear target value or target range, such as a contact angle of not less than 120° and a volume resistivity of not less than 1×10⁻⁶. 13 The required performance parameters are obtained by measuring Ω·cm and elongation at break of not less than 9%, thus obtaining the scenario-specific performance requirements expressed as performance parameters.

[0048] Starting with the basic performance of anti-condensation insulating coatings, a demand difference analysis is performed on the required performance for the aforementioned scenarios. To determine the target performance parameters, a basic formula is first screened from historical formula-performance correlation data. The screening rule is that the formula consistently meets the following requirements in multiple batches of samples: contact angle not less than 120°, volume resistivity not less than 1×10¹³ Ω·cm, and elongation at break not less than 9%. Then, the contact angle, volume resistivity, and elongation at break corresponding to the basic formula are used as basic performance values. Next, each performance parameter in the scenario's required performance is subtracted from the basic performance value to obtain the difference, and this difference is recorded in the demand difference record table. Finally, performance parameters with a difference of zero are marked as maintenance items, and performance parameters with a positive difference are marked as improvement items. The target values ​​or target ranges of the maintenance and improvement items are then summarized as the target performance parameters.

[0049] Finally, the target performance parameters are used as inputs to perform component performance response analysis through a component-performance prediction model. In this process, a candidate formulation set is first generated based on the initial component range. This set covers fixed values ​​of stable critical components, full-range variations of suspected critical components, fixed values ​​of stable auxiliary components, and narrow-range variations of suspected auxiliary components. Then, each candidate formulation is input into the component-performance prediction model to calculate the multi-objective performance prediction value for each formulation. These multi-objective performance prediction values ​​include anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental performance index. After obtaining the multi-objective performance prediction values, a weighted calculation is performed on the multi-objective performance prediction values ​​of each candidate formulation according to the priority weights in the target performance parameter set to obtain the corresponding comprehensive performance score. Formulations with comprehensive performance scores higher than a preset threshold are then selected. Candidate formulations are selected as the preferred candidate group. Next, Pareto front analysis is performed on the candidate formulations in the preferred candidate group to identify the non-dominated solution set that does not exhibit disadvantage in multiple performance indicators. From the non-dominated solution set, the formulations are sorted according to their matching degree with the target performance parameter set. The candidate formulation with the highest matching degree is selected as the first predictive theoretical component, the candidate formulation with the second highest matching degree but different component composition is selected as the second predictive theoretical component, and the candidate formulation located at the Pareto front boundary and exhibiting outstanding performance in a single performance indicator is selected as the third predictive theoretical component. These three components together constitute the predictive theoretical component set, which is then output, including detailed proportions of each predictive theoretical component, multi-objective performance prediction values, confidence assessments, and corresponding process condition recommendations.

[0050] Furthermore, in the method provided in the application embodiments, the target performance parameter is used as input, and component performance response analysis is performed through the component-performance prediction model to obtain the predicted theoretical component, which further includes:

[0051] Based on the initial range of components, a candidate formulation set is generated, covering fixed values ​​of stable key components, full-range variations of suspected key components, fixed values ​​of stable auxiliary components, and narrow-range variations of suspected auxiliary components. The candidate formulation set is input into the component-performance prediction model to calculate multi-objective performance prediction values ​​for each candidate formulation. These multi-objective performance prediction values ​​include anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental performance index. The multi-objective performance prediction values ​​are weighted according to the priority weights in the target performance parameter set to obtain a comprehensive performance score for each candidate formulation. Candidate formulations with comprehensive performance scores higher than a preset threshold are selected as preferred candidates. A Pareto front analysis is performed on each candidate formulation in the preferred candidate group to identify non-dominated solution sets that have no disadvantage in multiple performance indicators. From the non-dominated solution set, the candidate formulation with the highest matching degree with the target performance parameter set is selected as the first predictive theoretical component, the candidate formulation with the second highest matching degree and significant difference in component composition is selected as the second predictive theoretical component, and the candidate formulation at the Pareto front boundary with outstanding single performance is selected as the third predictive theoretical component, which together constitute the predictive theoretical component set. The predictive theoretical component set is output, which includes the detailed proportions of each predictive theoretical component, the predicted performance indicators, the confidence level assessment, and the process conditions required to achieve the predicted performance.

[0052] In this embodiment, when generating a candidate formulation set based on the initial range of components, the value rules for each component are first clarified. The stable key component is assigned a fixed value within its initial range, and the stable auxiliary component is assigned a fixed value. The suspected key component is divided into five value points within its entire range using an equidistant division method: the lower boundary value, the 25% position value, the 50% position value, the 75% position value, and the upper boundary value. The suspected auxiliary component is divided into three value points within its narrow range using an equidistant division method: the lower boundary value, the interval center value, and the upper boundary value. Then, a full combination enumeration method is used to perform Cartesian combinations of the component value points. Each combination forms a candidate formulation record containing the specific numerical proportions of the matrix resin, modified filler, functional additives, and curing system. All combination results are summarized to obtain a candidate formulation set covering the fixed values ​​of the stable key component, the entire range of the suspected key component, the fixed values ​​of the stable auxiliary component, and the narrow range of the suspected auxiliary component.

[0053] Next, the candidate formulation set is input into the component-performance prediction model. For each candidate formulation in the set, an input feature vector is constructed according to the order of the model input fields and input into the component-performance prediction model to obtain the corresponding basic performance prediction results, including predicted values ​​for volume resistivity, volume resistivity after aging, contact angle, contact angle after aging, elongation at break, and process-related prediction parameters. Subsequently, the above basic performance prediction results are converted into multi-objective performance prediction values. That is, for each basic performance index, linear normalization is performed based on the historical minimum and maximum values ​​of that index in the historical formulation-performance correlation data, where the normalized value is equal to the predicted value minus the historical minimum value. Then divide by the difference between the historical maximum and the historical minimum values; for the predicted volume resistivity and the predicted volume resistivity after aging, first perform logarithmic transformation and then perform linear normalization; after obtaining the normalized values ​​of each index, the anti-condensation performance index is equal to the arithmetic mean of the normalized contact angle value and the normalized contact angle value after aging, the electrical insulation performance index is equal to the arithmetic mean of the normalized volume resistivity value and the normalized volume resistivity value after aging, the durability performance index is equal to the normalized elongation at break value, and the process environmental performance index is equal to the normalized value of the process-related predicted parameters, thus obtaining the anti-condensation performance index, electrical insulation performance index, durability index, and process environmental performance index corresponding to each candidate formulation.

[0054] Then, based on the priority weights in the target performance parameter set, when calculating the weighted values ​​of the multi-target performance predictions, the priority weights are given in a preset manner and the sum of all priority weights is equal to 1. For each candidate formulation, the anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental protection performance index are multiplied by the corresponding priority weights, and the product results are added together to obtain the comprehensive performance score of the candidate formulation, thus obtaining the comprehensive performance score of each candidate formulation.

[0055] Subsequently, when selecting candidate formulations with comprehensive performance scores higher than a preset threshold as the preferred candidate group, the comprehensive performance scores of all candidate formulations are first sorted from largest to smallest, and the values ​​corresponding to the comprehensive performance scores in the top 10% after sorting are determined as the preset threshold; then, the comprehensive performance scores of candidate formulations are compared one by one, and candidate formulations with comprehensive performance scores higher than the preset threshold are retained to obtain the preferred candidate group.

[0056] When performing Pareto front analysis on the preferred candidate groups, any two candidate formulations are compared across four dimensions: anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental performance index. If there exists a candidate formulation B that satisfies the condition that the value of each of the four performance indices is greater than or equal to the corresponding value of candidate formulation A, and that the difference in value of at least one performance index is not less than 0.01, then candidate formulation A is determined to be a dominated solution and is eliminated. After completing all pairwise comparisons, the remaining candidate formulations constitute the set of non-dominated solutions.

[0057] When determining the predicted theoretical components from the non-dominated solution set, the weighted absolute difference between each candidate formulation and the target performance parameter set is calculated, and the candidate formulations are sorted from smallest to largest according to the weighted absolute difference. The candidate formulation with the largest weighted absolute difference is determined as the first predicted theoretical component. Among the remaining candidate formulations, the candidate formulation with the second smallest weighted absolute difference and at least one component content difference not less than 10% of the initial range width of that component is determined as the second predicted theoretical component. Among the non-dominated solution set, the candidate formulation that reaches the maximum value in the set on any performance index is determined as the third predicted theoretical component. The first predicted theoretical component, the second predicted theoretical component, and the third predicted theoretical component together constitute the predicted theoretical component set.

[0058] When outputting the set of predicted theoretical components, the system provides detailed proportions of the matrix resin, modified filler, functional additives, and curing system for each predicted theoretical component. It also outputs the corresponding anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental performance index as predicted performance indicators. The confidence assessment outputs the prediction accuracy value obtained from the component-performance prediction model in cross-validation, and records the number of samples in the historical formulation-performance correlation data for the component's value range. Simultaneously, it outputs suggested process conditions for achieving the predicted performance, including the mixing temperature range, stirring time, and curing time range, enabling the predicted theoretical component to be directly verifiable.

[0059] Step S400: Based on the target performance parameters and the predicted theoretical components and initial components, perform reliability deviation peak analysis to locate and test the verification components.

[0060] In this embodiment, when performing reliability deviation peak analysis based on the target performance parameters and the predicted theoretical components and initial components, the predicted theoretical components and initial components are first compared item by item to identify components with differences in values ​​as deviation components. Then, in conjunction with the search and determination process of the predicted theoretical components, the key evolution links corresponding to each deviation component during the optimization iteration process are extracted. These key evolution links characterize the continuous correlation path between changes in component content and changes in target performance parameters. Subsequently, test response relationship analysis is performed based on the key evolution links, and a structured analysis is conducted on the trend of target performance parameter changes corresponding to deviation components at different value nodes to determine the test verification component interval. After obtaining the test verification component interval, the test verification component interval is cross-screened with suspicious components to finally locate the test verification component.

[0061] Furthermore, in the method provided in the application embodiments, the method further includes performing reliability deviation peak analysis based on the target performance parameters and the predicted theoretical components and initial components to locate the test verification components.

[0062] Based on the predicted theoretical components and the initial components, deviation components are obtained, and according to the search and determination process of the predicted theoretical components, the key evolution links for locating the suspected component types of the deviation components are identified; test response relationship analysis is performed based on the key evolution links to determine the test verification component interval; and the test verification component is located based on the test verification component interval and the suspected components.

[0063] In the embodiments of this application, when obtaining the deviation component based on the predicted theoretical component and the initial component, the value of the predicted theoretical component and the initial component in each component of the matrix resin, modified filler, functional additive, and curing system is first compared using the item-by-item difference calculation method. That is, the value of the component in the predicted theoretical component is subtracted from the fixed value or interval center value of the component in the initial component, and the component with an absolute difference value greater than 5% of the initial range width of the component is marked as the deviation component.

[0064] Next, based on the search and determination process of the predicted theoretical components, the key evolution links for identifying the suspected component types of the deviation components are determined. In this process, iterative trajectory data of the optimization algorithm is first extracted from the search and determination process of the predicted theoretical components. This iterative trajectory data includes changes in the content of candidate solution components, changes in the overall performance score, and records of search direction adjustments in each iteration. Then, for each deviation component, its content evolution path in the iterative trajectory data is back-analyzed to identify abrupt changes in the content change rate, reversals in the search direction, and points of stagnation in multiple iterations. These key nodes are then connected in series to form the key evolution links for the deviation components.

[0065] Next, the test response relationship was analyzed based on the search for key evolution links. In this process, firstly, rate abrupt change points and direction reversal points located within the highly sensitive core region were identified in the search for key evolution links. Then, high-sensitivity segments were constructed by expanding outwards from these points. Subsequently, a structured analysis was performed on the correspondence between component content and target performance parameters within the high-sensitivity segments to confirm the existence of a strong causal relationship within this interval, and to identify the nonlinear characteristics of the performance response, including abrupt change thresholds, extreme points, and sensitive windows. Finally, using the rate abrupt change points and direction reversal points within the high-sensitivity segments as anchor points, and combining this with the nonlinear characteristics of the performance response within the interval, the test verification component interval was determined.

[0066] Finally, when locating test verification components based on the test verification component intervals and suspicious components, the following steps are first taken: determining the deviation component set, the suspicious component set, and the component set with test verification component intervals. The deviation component set consists of components whose absolute difference in content between the predicted theoretical component and the initial component is greater than 5% of the initial range width. The suspicious component set consists of components marked as suspicious key components or suspicious auxiliary components in the component type classification results. The component set with test verification component intervals consists of components whose highly sensitive segments are identified and clearly defined upper and lower boundary intervals are formed during the test response relationship analysis process. Subsequently, each component in the deviation component set... Each component is evaluated item by item. First, it is determined whether it belongs to the suspicious component set. If not, it is removed. If it belongs, it is further determined whether there is a corresponding test verification component interval. The test verification component interval is checked to see if it meets the conditions that the lower boundary of the interval is greater than or equal to the lower boundary of the initial range of the component, the upper boundary of the interval is less than or equal to the upper boundary of the initial range of the component, and the interval width is greater than zero. When all three conditions are met simultaneously, namely, that it is a deviation component, belongs to the suspicious component type, and has a valid test verification component interval, the component is determined as a test verification component, and its test verification component interval is recorded as the boundary for subsequent measured values, thereby completing the location of the test verification component.

[0067] Furthermore, in the method provided in the application embodiments, the search key evolution link for locating the suspected component type of the deviation component according to the search and determination process of the predicted theoretical component further includes:

[0068] Based on the search and determination process of the predicted theoretical components, the iterative trajectory data of the optimization algorithm in finding the optimal solution is extracted. The iterative trajectory data includes the component content changes, comprehensive performance score changes, and search direction adjustment records of the candidate solutions in each iteration. For each deviation component, the content evolution path in the iterative trajectory is traced, and the abrupt change point of the content change rate, the search direction reversal point, and the stagnation point of multiple iterations are marked to obtain the key evolution link of the deviation component search.

[0069] In this embodiment, when extracting the iterative trajectory data of the optimization algorithm in finding the optimal solution based on the search and determination process of the predicted theoretical components, an iterative log recording method is used to collect data throughout the entire optimization algorithm process. After the optimization algorithm is initialized, the current candidate solution is written to the iterative log file at the end of each iteration. The iterative log file includes the iteration round number, matrix resin content, modified filler content, functional additive content, curing system content, comprehensive performance score, and search direction adjustment record. The search direction adjustment record is used to identify whether the adjustment direction of each component in the current round is an increase or a decrease compared to the previous round. After the optimization algorithm finishes running, the iterative log file is read and sorted in ascending order by the iteration round number to form a complete iterative trajectory data sequence, thereby obtaining the iterative trajectory data of the optimization algorithm in the process of finding the optimal solution.

[0070] When tracing the content evolution path of each deviation component in the iterative trajectory, a time series extraction method is used to extract the content values ​​of the deviation component in each iteration from the iterative trajectory data sequence. A time series of the deviation component content is constructed according to the iteration round, and the comprehensive performance score time series for the corresponding round is also extracted. Subsequently, a difference calculation method is used to perform a first-order difference operation on the deviation component content time series to obtain the content change rate sequence between adjacent iterations. The mean and standard deviation of this content change rate sequence are calculated. When the absolute value of the content change rate in a certain round is greater than the mean plus twice the standard deviation, that round is marked as a point of abrupt change in the content change rate. Simultaneously, based on the search... The direction adjustment record uses symbol encoding to represent the adjustment direction of the deviation component. When the direction sign changes from positive to negative or from negative to positive in two consecutive rounds, that round is marked as a search direction reversal point. In addition, a sliding window statistical method is used to mark the consecutive round segments as multi-round iteration stagnation points when the change amplitude of the deviation component content is less than 1% of the initial range width of the component in three or more consecutive iterations. Finally, the abrupt change points of content change rate, search direction reversal points, and multi-round iteration stagnation points are integrated according to the iteration sequence to form a search key evolution link containing key node numbers and corresponding comprehensive performance score change data, thereby obtaining the search key evolution link of the deviation component.

[0071] Furthermore, in the method provided in the application embodiments, the process of parsing test response relationships based on the search key evolution links to determine the test verification component intervals also includes:

[0072] Identify rate abrupt change points and direction reversal points located in the highly sensitive core region within the key evolution link of the search, and expand outwards from the key points to establish a complete highly sensitive segment; obtain a strong causal relationship between the component content and target performance parameters within the highly sensitive segment, and identify the nonlinear characteristics of the performance response within the interval, including abrupt change thresholds, extreme points, or sensitive windows; use the rate abrupt change points and direction reversal points within the highly sensitive segment as anchor points, and combine the nonlinear characteristics of the performance response within the interval to locate the test and verification component interval.

[0073] In this embodiment, when identifying rate abrupt change points and direction reversal points located in the highly sensitive core region during the search of critical evolution links, the component content values ​​of the deviation components in rounds 1 to n, and the target performance parameter values ​​of the corresponding rounds, are first extracted according to the iteration order. The component content values ​​between adjacent rounds are subtracted to obtain the component content difference between adjacent rounds. At the same time, the target performance parameter values ​​between adjacent rounds are subtracted to obtain the performance difference between adjacent rounds. Under the premise that the component content difference is not zero, the performance difference is divided by the component content difference to obtain the performance change ratio caused by the change of a unit component. Then, the average and standard deviation of all performance change ratios are calculated. When the absolute value of a certain performance change ratio is greater than the average plus twice the standard deviation, the position is marked as a rate abrupt change point. At the same time, the sign of the component content difference between two consecutive rounds is judged. When the sign of the component content difference changes from positive to negative or from negative to positive, the corresponding position is marked as a direction reversal point. The continuous iteration segment where the rate abrupt change point and the direction reversal point are located is determined as the highly sensitive core region.

[0074] When establishing a complete high-sensitivity segment by expanding outwards from the rate change point and direction reversal point, the expansion proceeds in iterations, starting from each rate change point or direction reversal point and proceeding forward and backward in sequence. The expansion rule is to recalculate the performance change ratio of the newly added segment. If the absolute value of the performance change ratio of the newly added segment is still greater than the average value, the expansion continues. If the absolute value of the performance change ratio is less than or equal to the average value, the expansion stops. After expanding all key points, the overlapping expanded segments are merged to form one or more continuous segments, and each continuous segment is a high-sensitivity segment.

[0075] To identify a strong causal relationship between component content and target performance parameters within a highly sensitive range and to recognize the nonlinear characteristics of performance response within that range, the component content values ​​within the highly sensitive range are first reordered from smallest to largest, and the corresponding target performance parameter values ​​are simultaneously arranged. The component content difference and performance difference between adjacent points are calculated segment by segment, and the corresponding performance change ratio is calculated. When the absolute value of the performance change ratio for two or more consecutive segments near a certain position is greater than the average plus twice the standard deviation, the corresponding component content position is determined as the mutation threshold. When the target performance parameter value corresponding to a component content is simultaneously greater than or simultaneously less than its adjacent values, the component content position is determined as an extreme point. When the absolute value of the performance change ratio for each segment within a continuous component content range is greater than the average, the continuous range is determined as a sensitive window. This process identifies the nonlinear characteristics of performance response, such as mutation thresholds, extreme points, and sensitive windows, thus demonstrating a strong causal relationship between component content changes and target performance parameter changes within the highly sensitive range.

[0076] Finally, using the rate mutation points and direction reversal points within the highly sensitive segment as anchor points, the component content values ​​corresponding to all rate mutation points, direction reversal points, mutation thresholds, extreme points, and the start and end positions of the sensitive window are summarized. The minimum component content value is taken as the lower boundary of the test and verification component interval, and the maximum component content value is taken as the upper boundary of the test and verification component interval. This interval is then checked against the initial component range to ensure that the test and verification component interval is completely within the initial component range, thereby locating the test and verification component interval.

[0077] Step S500: Conduct experimental testing and verification based on the tested and verified components to determine the final optimized target coating components.

[0078] In this embodiment of the application, when conducting experimental testing and verification based on the test and verification components, test formulations numbered 1 to 8 are constructed within the test and verification component range, and control 1 to control 7 are set as comparative samples. After the sample preparation is completed under the conditions of uniform raw material weighing accuracy, mixing and stirring time, dispersion conditions and curing process, the target performance parameters of each formulation are tested, and the following specific data are obtained.

[0079] Among them, the volume resistivity of number 1 is 3.2 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 3.6×10⁻⁶. 14 The resistance value is Ω·cm, the contact angle is 125°, the contact angle after aging is 123°, the elongation at break is 9.5%, and the withstand voltage test is 42.06kV. The volume resistivity of item number 2 is 4.5×10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 4.4×10⁻⁶. 14The contact angle was 122° (Ω·cm), 121° after aging, and the elongation at break was 9.3%. The volume resistivity of item number 3 was 3.6 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 4.0×10⁻⁶. 14 Ω·cm, contact angle 122°, contact angle after aging 122°, elongation at break 9.7%. Number 4 has a volume resistivity of 3.8 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 4.1×10⁻⁶. 14 Ω·cm, contact angle 125°, contact angle after aging 120°, elongation at break 9.2%. Number 5 has a volume resistivity of 3.6 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 3.9×10 Ω·cm. 14 The contact angle was 122° (Ω·cm), 121° after aging, and the elongation at break was 9.7%. The volume resistivity of item number 6 was 3.0 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 3.3×10 Ω·cm. 14 The contact angle was 119° (Ω·cm), 115° after aging, and the elongation at break was 9.5%. The volume resistivity of item number 7 was 2.9 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 3.0×10⁻⁶. 14 The contact angle was 114° (Ω·cm), 112° after aging, and the elongation at break was 9.5%. The volume resistivity of item number 8 was 4.0 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 4.1×10⁻⁶. 14 The contact angle was 119° Ω·cm, and the contact angle after aging was 118°. The elongation at break was 4.3%.

[0080] The volume resistivity of reference 1 is 2.1 × 10⁻⁶. 14 The resistance to solidification was Ω·cm, the contact angle was 127°, and the elongation at break was 4.3%. The volume resistivity of control 2 was 1.9 × 10⁻⁶. 13 Ω·cm, the volume resistivity after aging is 9.3×10⁻⁶. 13 The contact angle was 121° (Ω·cm) after aging and 107° after aging. The elongation at break was 6.4%. The volume resistivity of control 3 was 1.6 × 10⁻⁶. 14 Ω·cm, the volume resistivity after aging is 2.3×10⁻⁶. 14 The contact angle was 112° (Ω·cm) after aging and 104° after aging. The elongation at break was 6.1%. The volume resistivity of control 4 was 2.5 × 10⁻⁶. 13 Ω·cm, the volume resistivity after aging is 1.8×10⁻⁶. 12The initial contact angle was 107°, which decreased to 96° after aging, and the elongation at break was 5.3%. The volume resistivity of control 5 was 8.4 × 10¹³ Ω·cm, and the volume resistivity after aging was 2.2 × 10¹³ Ω·cm. 12 The contact angle was 110° (Ω·cm), decreasing to 98° after aging, and the elongation at break was 4.6%. The volume resistivity of control 6 was 7.3 × 10⁻⁶. 13 Ω·cm, the volume resistivity after aging is 1.3×10⁻⁶. 12 The contact angle was 115° (Ω·cm), and after aging, it became 101°. The elongation at break was 5.7%. The volume resistivity of control 7 was 2.4 × 10⁻⁶. 12 Ω·cm, the volume resistivity after aging is 3.0×10⁻⁶. 12 Ω·cm, contact angle of 101°, contact angle of 101° after aging, and elongation at break of 9.1%.

[0081] The selection process is conducted item by item based on preset screening criteria, with the screening criterion being a volume resistivity of not less than 3.0 × 10⁻⁶. 14 Ω·cm, and the volume resistivity after aging is not less than 3.0×10⁻⁶. 14 Ω·cm, contact angle not less than 120°, contact angle after aging not less than 118°, elongation at break not less than 9.0%, withstand voltage test not less than 40kV.

[0082] Comparing the above conditions one by one, item 1 meets all the conditions, and its volume resistivity is 3.2 × 10⁻⁶. 14 Ω·cm, volume resistivity after aging is 3.6×10⁻⁶. 14 Ω·cm, contact angle 125°, contact angle after aging 123°, elongation at break 9.5%, withstand voltage test 42.06kV. Number 2 meets all conditions, with a volume resistivity of 4.5×10⁻⁶. 14 Ω·cm, volume resistivity after aging is 4.4×10⁻⁶. 14 Ω·cm, contact angle 122°, contact angle after aging 121°, elongation at break 9.3%. Number 3 meets all conditions, with a volume resistivity of 3.6 × 10⁻⁶. 14 Ω·cm, volume resistivity after aging is 4.0×10⁻⁶. 14 Ω·cm, contact angle 122°, contact angle after aging 122°, elongation at break 9.7%. Item 4 meets all conditions, with a volume resistivity of 3.8 × 10⁻⁶. 14 Ω·cm, volume resistivity after aging is 4.1×10⁻⁶. 14 Ω·cm, contact angle 125°, contact angle after aging 120°, elongation at break 9.2%. Item 5 meets all conditions, with a volume resistivity of 3.6 × 10⁻⁶. 14 Ω·cm, volume resistivity after aging is 3.9×10⁻⁶. 14Ω·cm, contact angle 122°, contact angle after aging 121°, elongation at break 9.7%.

[0083] Number 6 does not meet the condition because its contact angle of 119° is lower than 120°. Number 7 does not meet the condition because its volume resistivity is 2.9 × 10⁻⁶. 14 Ω·cm less than 3.0×10 14 The contact angle was 114°, which is less than 120°, thus failing to meet the requirements. Sample No. 8 failed to meet the requirements because its elongation at break was 4.3%, which is less than 9.0%. All control samples failed to meet the requirements because at least one performance criterion was lower than the screening criteria.

[0084] Among items numbered 1 to 5 that meet all screening criteria, the volume resistivity is compared with the volume resistivity after aging. Item number 2 has a volume resistivity of 4.5 × 10⁻⁶. 14 Ω·cm, volume resistivity after aging is 4.4×10⁻⁶. 14 Ω·cm, the highest among the five groups; volume resistivity of group 3 is 3.6×10. 14 Ω·cm, volume resistivity after aging is 4.0×10⁻⁶. 14 Ω·cm; Volume resistivity of No. 4: 3.8 × 10⁻⁶ 14 Ω·cm, volume resistivity after aging is 4.1×10⁻⁶. 14 Ω·cm; Volume resistivity of No. 5: 3.6 × 10⁻⁶ 14 Ω·cm, volume resistivity after aging is 3.9×10⁻⁶. 14 Ω·cm; Although the volume resistivity of No. 1 is slightly lower than that of No. 2, it has a withstand voltage test result of 42.06kV.

[0085] Based on the combined results of absolute volume resistivity, volume resistivity after aging, and elongation at break, and provided all screening criteria are met, candidates with both volume resistivity and volume resistivity after aging reaching 4.0 × 10⁻⁶ were selected. 14 The component combination with an Ω·cm or higher and an elongation at break of not less than 9.3% was determined as the final optimized target coating component, and the final optimized target coating component was verified through experimental testing.

[0086] In summary, by collecting physicochemical parameter data of the matrix resin, modified filler, functional additives, and curing system, as well as historical formulation-performance correlation data, a component database for anti-condensation insulating coatings was established. This database establishes a statistically analyzable correspondence between the content of each component and target performance parameters such as volume resistivity, contact angle, elongation at break, and withstand voltage, providing a data foundation for subsequent quantitative analysis. Based on this data, the initial range of components was determined, and a component-performance prediction model was established. This model allows for the prediction of the impact of component ratio changes on multiple performance indicators, enabling the comparison and screening of performance levels of different ratios even before actual sample preparation. By performing multi-objective performance prediction value calculations, weighted scoring, and Pareto front analysis on the candidate formulation set, the interrelationships between multiple performance indicators are transformed into comparable comprehensive evaluation results, thereby screening out the theoretically predicted components that are not inferior in multiple performance aspects. Furthermore, by analyzing the deviation between the theoretically predicted components and the initial components, the test verification component range is located, ensuring that the experimental verification focuses on the component range that has a significant impact on performance and is subject to uncertainty, avoiding repeated adjustments to stable components and reducing ineffective experiments. Based on the above optimization process, which involves data support, model prediction, comprehensive screening, and targeted verification leading to convergence, it was finally determined that the volume resistivity must simultaneously meet the following conditions: not less than 3.0 × 10⁻⁶. 14 Ω·cm, volume resistivity after aging is not less than 3.0×10 14 The optimized target coating components meet the requirements of Ω·cm, contact angle not less than 120°, contact angle after aging not less than 118°, elongation at break not less than 9.0%, and withstand voltage not less than 40kV. This achieves a synergistic improvement in anti-condensation performance, electrical insulation performance, and durability performance, while also improving the efficiency and accuracy of formulation optimization and prediction, and shortening the R&D cycle.

[0087] In summary, the embodiments of this application have at least the following technical effects:

[0088] This application collects physicochemical parameter data of the base resin, modified filler, functional additives, and curing system, as well as historical formulation-performance correlation data, to establish a component database for anti-condensation insulating coatings. Based on the component database, an initial range of components is determined and a component-performance prediction model is established. Using the target performance parameters analyzed in the application scenario as the objective, the initial range of components is optimized through the component-performance prediction model to obtain predicted theoretical components. Based on the target performance parameters and the components of the predicted theoretical components and the initial components, a reliability deviation peak analysis is performed to locate the test verification components. Based on the test verification components, experimental testing and verification are conducted to determine the final optimized target coating components. This invention solves the technical problem in the prior art where anti-condensation insulating coating formulation optimization relies on manual experience and is difficult to achieve synergistic optimization of multiple performance indicators. By establishing a component database and constructing a component-performance prediction model for model-driven optimization analysis, it achieves intelligent synergistic optimization of multiple target performances and improves the efficiency and accuracy of formulation optimization.

[0089] Example 2, based on the same inventive concept as the intelligent optimization method for the components of an anti-condensation insulating coating in the foregoing examples, such as... Figure 2 As shown, this application provides a smart optimization system for the components of an anti-condensation insulating coating. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0090] The database establishment module 11 is used to collect physicochemical parameter data of the base resin, modified filler, functional additives, and curing system, as well as historical formula-performance correlation data, to establish a component database for anti-condensation insulating coatings; the model establishment module 12 is used to determine the initial range of components and establish a component-performance prediction model based on the component database; the optimization module 13 is used to optimize the initial range of components using the component-performance prediction model with the target performance parameters analyzed in the application scenario as the target, to obtain the predicted theoretical components; the analysis module 14 is used to perform reliability deviation peak analysis based on the target performance parameters and the predicted theoretical components and initial components, to locate the test verification components; the test verification module 15 is used to conduct experimental tests and verifications based on the test verification components to determine the final optimized target coating components.

[0091] Furthermore, the system is also used to implement the following functions:

[0092] Based on existing research and experimental datasets, the influence of each component on coating performance and the performance fluctuation of the components in actual production are analyzed. The components are classified into types, including stable key components with large performance impact and small fluctuation, suspected key components with large performance impact and large fluctuation, stable auxiliary components with small performance impact and small fluctuation, and suspected auxiliary components with small performance impact and large fluctuation. According to the influence relationship between each component type and performance parameters, the component types are mapped and associated with target performance parameters to establish a component database for anti-condensation insulating coatings.

[0093] Furthermore, the system is also used to implement the following functions:

[0094] Obtain the statistical distribution of the content of each component in the existing formulas stored in the component database, and determine the feasible range of each component by combining physicochemical constraints and expert experience; based on the feasible range of each component, screen the fixed values ​​or ranges of the performance of each component with the basic performance of the anti-condensation insulating coating as the benchmark, and obtain the fixed values ​​or ranges of each component; determine the initial range of the component according to the fixed values ​​or ranges of each component.

[0095] Furthermore, the system is also used to implement the following functions:

[0096] Using historical formula data and historical formula-performance correlation data from the component database as training samples, with component content as input features and target performance parameters as output labels, an initial prediction model is trained using a machine learning algorithm. The prediction accuracy of the initial prediction model is evaluated using cross-validation. When the prediction accuracy does not meet the preset standard, sparse data regions are analyzed, experimental points are added to the sparse data regions, and the training samples are expanded using the supplemented experimental data. The model is then retrained until the prediction accuracy meets the preset standard, thus obtaining the component-performance prediction model.

[0097] Furthermore, the system is also used to implement the following functions:

[0098] Based on the application scenario, the anti-condensation insulation requirements are analyzed to determine the required performance for the scenario. Starting from the basic performance of the anti-condensation insulating coating, the required performance for the scenario is analyzed to determine the target performance parameters. Using the target performance parameters as input, the component performance response is analyzed through the component-performance prediction model to obtain the predicted theoretical components.

[0099] Furthermore, the system is also used to implement the following functions:

[0100] Based on the initial range of components, a candidate formulation set is generated, covering fixed values ​​of stable key components, full-range variations of suspected key components, fixed values ​​of stable auxiliary components, and narrow-range variations of suspected auxiliary components. The candidate formulation set is input into the component-performance prediction model to calculate multi-objective performance prediction values ​​for each candidate formulation. These multi-objective performance prediction values ​​include anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental performance index. The multi-objective performance prediction values ​​are weighted according to the priority weights in the target performance parameter set to obtain a comprehensive performance score for each candidate formulation. Candidate formulations with comprehensive performance scores higher than a preset threshold are selected as preferred candidates. A Pareto front analysis is performed on each candidate formulation in the preferred candidate group to identify non-dominated solution sets that have no disadvantage in multiple performance indicators. From the non-dominated solution set, the candidate formulation with the highest matching degree with the target performance parameter set is selected as the first predictive theoretical component, the candidate formulation with the second highest matching degree and significant difference in component composition is selected as the second predictive theoretical component, and the candidate formulation at the Pareto front boundary with outstanding single performance is selected as the third predictive theoretical component, which together constitute the predictive theoretical component set. The predictive theoretical component set is output, which includes the detailed proportions of each predictive theoretical component, the predicted performance indicators, the confidence level assessment, and the process conditions required to achieve the predicted performance.

[0101] Furthermore, the system is also used to implement the following functions:

[0102] Based on the predicted theoretical components and the initial components, deviation components are obtained, and according to the search and determination process of the predicted theoretical components, the key evolution links for locating the suspected component types of the deviation components are identified; test response relationship analysis is performed based on the key evolution links to determine the test verification component interval; and the test verification component is located based on the test verification component interval and the suspected components.

[0103] Furthermore, the system is also used to implement the following functions:

[0104] Based on the search and determination process of the predicted theoretical components, the iterative trajectory data of the optimization algorithm in finding the optimal solution is extracted. The iterative trajectory data includes the component content changes, comprehensive performance score changes, and search direction adjustment records of the candidate solutions in each iteration. For each deviation component, the content evolution path in the iterative trajectory is traced, and the abrupt change point of the content change rate, the search direction reversal point, and the stagnation point of multiple iterations are marked to obtain the key evolution link of the deviation component search.

[0105] Furthermore, the system is also used to implement the following functions:

[0106] Identify rate abrupt change points and direction reversal points located in the highly sensitive core region within the key evolution link of the search, and expand outwards from the key points to establish a complete highly sensitive segment; obtain a strong causal relationship between the component content and target performance parameters within the highly sensitive segment, and identify the nonlinear characteristics of the performance response within the interval, including abrupt change thresholds, extreme points, or sensitive windows; use the rate abrupt change points and direction reversal points within the highly sensitive segment as anchor points, and combine the nonlinear characteristics of the performance response within the interval to locate the test and verification component interval.

[0107] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0108] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligently optimizing the components of an anti-condensation insulating coating, characterized in that, include: Collect physicochemical parameter data of matrix resin, modified filler, functional additives, and curing system, as well as historical formula-performance correlation data, to establish a component database for anti-condensation insulating coatings; Based on the component database, the initial range of components is determined and a component-performance prediction model is established; Using the target performance parameters analyzed in the application scenario as the objective, the initial range of the components is optimized through the component-performance prediction model to obtain the predicted theoretical components; Based on the target performance parameters and the predicted theoretical components and initial components, a reliability deviation peak analysis is performed to locate and test the verification components. Based on the test and verification components, conduct experimental tests and verifications to determine the final optimized target coating components; Specifically, based on the target performance parameters and the predicted theoretical components and initial components, a reliability deviation peak analysis is performed to locate the test verification components, including: Based on the predicted theoretical components and the initial components, the deviation components are obtained, and according to the search and determination process of the predicted theoretical components, the key evolution links for the search of suspicious component types are located for the deviation components. Based on the search key evolution link, test response relationship analysis is performed to determine the test verification component interval; Based on the test verification component range and the suspected components, locate the test verification component; The process of searching and determining the predicted theoretical components, including locating the key evolutionary links of the deviation components to identify the suspected component types, comprises: Based on the search and determination process of the predicted theoretical components, the iterative trajectory data of the optimization algorithm in finding the optimal solution is extracted. The iterative trajectory data includes the changes in the component content of the candidate solution, the changes in the comprehensive performance score, and the search direction adjustment records in each iteration. For each deviation component, trace the content evolution path in the iteration trajectory, mark the abrupt change point of content change rate, the reversal point of search direction, and the stagnation point of multiple iterations, and obtain the key evolution link of the deviation component search. Specifically, based on the search key evolution link, test response relationship parsing is performed to determine the test verification component interval, including: Identify rate abrupt changes and direction reversal points located in the highly sensitive core area of ​​the key evolution link, and expand outwards from the key points to establish a complete highly sensitive segment. The study aims to establish a strong causal relationship between the component content and target performance parameters within a highly sensitive range, and to identify the nonlinear characteristics of the performance response within the range, including abrupt change thresholds, extreme points, or sensitive windows. Using the rate abrupt change point and direction reversal point within the highly sensitive segment as anchor points, and combining the nonlinear characteristics of the performance response within the interval, the test and verification component interval is located.

2. The intelligent optimization method for the composition of the anti-condensation insulating coating according to claim 1, characterized in that, Establish a component database for anti-condensation insulating coatings, including: Based on existing research and experimental datasets, the influence of each component on coating performance and the performance fluctuation of the components in actual production are analyzed. The components are classified into types, including stable key components with large performance impact and small fluctuation, suspected key components with large performance impact and large fluctuation, stable auxiliary components with small performance impact and small fluctuation, and suspected auxiliary components with small performance impact and large fluctuation. Based on the influence relationship between each component type and performance parameters, the component types are mapped and associated with the target performance parameters to establish a component database for anti-condensation insulating coatings.

3. The intelligent optimization method for the composition of the anti-condensation insulating coating according to claim 2, characterized in that, Based on the component database, the initial range of components is determined, including: Obtain the statistical distribution of the content of each component in the existing formulas stored in the component database, and determine the feasible range for each component by combining physicochemical constraints and expert experience. Based on the feasible range of each component, the fixed values ​​or ranges of the performance of each component are screened using the basic performance of the anti-condensation insulating coating as a benchmark, so as to obtain the fixed values ​​or ranges of each component. The initial range of the components is determined based on the fixed values ​​or ranges of each component.

4. The intelligent optimization method for the composition of the anti-condensation insulating coating according to claim 2, characterized in that, Establish a component-performance prediction model, including: Using historical formula data and historical formula-performance correlation data in the component database as training samples, with component content as input feature and target performance parameter as output label, an initial prediction model is trained using a machine learning algorithm. The prediction accuracy of the initial prediction model is evaluated using cross-validation. When the prediction accuracy does not meet the preset standard, the sparse data regions are analyzed, experimental points are added to the sparse data regions, and the training samples are expanded using the supplemented experimental data. The model is then retrained until the prediction accuracy meets the preset standard, thus obtaining the component-performance prediction model.

5. The intelligent optimization method for the composition of the anti-condensation insulating coating according to claim 4, characterized in that, Using the target performance parameters analyzed in the application scenario as the objective, the initial range of the components is optimized through the component-performance prediction model to obtain the predicted theoretical components, including: Based on the application scenario, analyze the anti-condensation insulation requirements and determine the required performance for the scenario. Starting with the basic performance of anti-condensation insulating coatings, a demand difference analysis is performed on the required performance of the scenario to determine the target performance parameters. Using the target performance parameters as input, the component performance response is analyzed through the component-performance prediction model to obtain the predicted theoretical components.

6. The intelligent optimization method for the composition of the anti-condensation insulating coating according to claim 5, characterized in that, Using the target performance parameters as input, the component performance response is analyzed through the component-performance prediction model to obtain the predicted theoretical components, including: Based on the initial range of the components, a set of candidate formulations is generated, which covers the fixed values ​​of stable key components, the full range of variations of suspected key components, the fixed values ​​of stable auxiliary components, and the narrow range of variations of suspected auxiliary components. The candidate formulation set is input into the component-performance prediction model to calculate the multi-objective performance prediction value of each candidate formulation. The multi-objective performance prediction value includes the anti-condensation performance index, electrical insulation performance index, durability performance index, and process environmental protection performance index. Based on the priority weights in the set of target performance parameters, the multi-target performance prediction values ​​are weighted and calculated to obtain the comprehensive performance score of each candidate formulation. Candidate formulations with comprehensive performance scores higher than a preset threshold are selected as the preferred candidate group. Pareto front analysis is performed on each candidate formulation in the preferred candidate group to identify the nondominated solution set that has no disadvantage in multiple performance indicators. From the set of non-dominated solutions, the candidate formulation with the highest matching degree with the set of target performance parameters is selected as the first prediction theory component, the candidate formulation with the second highest matching degree and significant differences in component composition is selected as the second prediction theory component, and the candidate formulation located at the Pareto front boundary and with outstanding single performance is selected as the third prediction theory component, together forming the prediction theory component set. Output the set of predicted theoretical components, which includes detailed proportions of each predicted theoretical component, predicted performance indicators, confidence level assessments, and recommendations for process conditions required to achieve the predicted performance.

7. A component intelligent optimization system for an anti-condensation insulating coating, characterized in that, The system is used to execute the intelligent component optimization method for an anti-condensation insulating coating as described in any one of claims 1-6, and the system includes: The database creation module is used to collect physicochemical parameter data of matrix resin, modified filler, functional additives, and curing system, as well as historical formula-performance correlation data, to establish a component database for anti-condensation insulating coatings. The model building module is used to determine the initial range of components and build a component-performance prediction model based on the component database. The optimization module is used to optimize the initial range of the components using the component-performance prediction model, with the target performance parameters analyzed in the application scenario as the objective, to obtain the predicted theoretical components. The analysis module is used to perform reliability deviation peak analysis based on the target performance parameters and the predicted theoretical components and initial components, and to locate the test verification components. The testing and verification module is used to conduct experimental testing and verification based on the testing and verification components to determine the final optimized target coating components.

Images