Performance prediction method of nano cerium dioxide ultraviolet shielding coating based on machine learning optimization

Abstract

The invention discloses a method for predicting the performance of a nano cerium dioxide ultraviolet shielding coating based on machine learning optimization. The method comprises the following steps: preparing and testing the performance of a cerium dioxide ultraviolet shielding coating by an experimental means to obtain data, selecting key characteristic parameters from the data, and constructing a database between the key characteristic parameters and the performance of the ultraviolet shielding coating; and constructing a machine learning model by adopting a machine learning method, training the model by utilizing data in the database, and selecting an optimal model. During application, cumulative particle size distribution of nano cerium dioxide, a surface modifier, an organic solvent, the type of a fluorocarbon coating, a ratio of the fluorocarbon coating to nano cerium dioxide, and light transmittance of a cerium dioxide dispersion liquid which form a coating to be detected are used as input and are input into a selected model; the ultraviolet shielding performance and the visible light transmission performance of the coated ultraviolet shielding coating can be predicted. Therefore, a large number of screening experiments are omitted, and time and economic cost are saved.

Description

Technical Field

[0001] The present invention belongs to the technical field of nanomaterials, and in particular relates to a performance prediction method of a nano-cerium dioxide ultraviolet shielding coating based on machine learning optimization. Background Art

[0002] In recent years, with the destruction of the ozone layer, the amount of ultraviolet (UV) radiation has increased dramatically. Long-term exposure to ultraviolet radiation will have many adverse effects on human production and life. Therefore, it is necessary to develop materials that can shield ultraviolet rays and reduce the harm of ultraviolet rays, especially transparent ultraviolet shielding materials that can block most ultraviolet rays and highly transmit visible light. They can be added to the protected materials or coated on the surface of the materials, and achieve the purpose of protection through multiple shielding effects such as absorption, reflection, and scattering of ultraviolet light.

[0003] Traditional organic UV absorbers have been widely used, but they are prone to self-consumption, resulting in poor long-term stability. Therefore, inorganic metal oxides that can reflect, scatter or absorb UV rays while not absorbing visible light are suitable choices. Ceria with a wide band gap has the characteristics of high physical and chemical stability, non-toxicity, and low price. At the same time, it has low photocatalytic activity and will not produce too much active oxygen after absorbing UV rays to damage the substrate. Therefore, when choosing cerium dioxide as an inorganic UV shielding agent, it is necessary to control the particle size at the nanometer level and modify the particle surface to improve its compatibility with the substrate.

[0004] Nano-cerium dioxide ultraviolet shielding coating is a composite coating comprising nano-cerium dioxide, organic solvents and fluorocarbon coatings. The composite coating is coated on the surface of the material and subjected to heat treatment to obtain the ultraviolet shielding coating. The performance of the ultraviolet shielding coating (such as ultraviolet shielding performance and visible light transmittance) is closely related to the selection of nano-cerium dioxide, organic solvents, fluorocarbon coatings, etc. in the composite coating. To obtain a composite coating with performance that meets the requirements or a coating that meets the performance requirements often requires a large number of screening experiments, which will consume too much manpower and material costs. The present invention intends to provide certain predictions and guidance for the experiment through machine learning methods, thereby saving time and economic costs. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a performance prediction method for nano-cerium dioxide ultraviolet shielding coatings based on machine learning optimization.

[0006] The performance prediction method of nano-cerium dioxide ultraviolet shielding coating based on machine learning optimization provided by the present invention comprises the following steps:

[0007] S1, preparing a precursor solution for preparing nano-cerium dioxide particles, adding a precipitant to synthesize nano-cerium dioxide precipitates by a precipitation method, centrifuging and washing the precipitates to obtain nano-cerium dioxide particles; changing the preparation conditions to prepare multiple groups of different nano-cerium dioxide particles;

[0008] S2, adding a surface modifier to each group of nano-cerium dioxide and dispersing it in an organic solvent to obtain nano-cerium dioxide particles with good dispersibility, mixing the organic solvent dispersion of nano-cerium dioxide with a fluorocarbon coating to obtain a composite coating, coating with the composite coating and preparing multiple groups of transparent coatings with nano-cerium dioxide through heat treatment, and testing the ultraviolet shielding performance and visible light transmittance of the coatings;

[0009] S3, taking the cumulative particle size distribution of nano-cerium dioxide, the surface modifier in step S2, the organic solvent, the light transmittance of the dispersion, the type of fluorocarbon coating, and the ratio of fluorocarbon coating to nano-cerium dioxide as characteristic values, and the ultraviolet shielding performance and visible light transmittance of the coating as target variables, to establish a database;

[0010] S4 builds machine learning models based on different machine learning methods, and uses the S3 database to train and cross-validate the machine learning models to obtain trained machine learning models; uses the scoring function to evaluate the machine learning models and selects the model with the best fitting effect.

[0011] S5. For the nano-cerium dioxide ultraviolet shielding coating that needs to be predicted for performance, the cumulative particle size distribution of the nano-cerium dioxide, the type of surface modifier, organic solvent, fluorocarbon coating, the ratio of fluorocarbon coating and nano-cerium dioxide, and the transmittance of the cerium dioxide dispersion are taken as input and input into the model selected in S4. This can predict the ultraviolet shielding performance and visible light transmittance of the ultraviolet shielding coating applied thereon.

[0012] Preferably, in S4, the machine learning method is one or more of linear regression, support vector machine, random forest regression, ridge regression, and extreme gradient boosting method.

[0013] Preferably, the cross-validation uses k-fold cross-validation, which divides the data set into multiple subsets and uses different combinations of training sets and test sets to train and validate the model multiple times, thereby effectively evaluating the performance of the model on different data, reducing the risk of overfitting, and enhancing the generalization ability of the model.

[0014] Preferably, the score function is the determination coefficient R 2The root mean square error (RMSE) measures the accuracy of the model's predictions, and the closer it is to 0, the smaller the prediction error.

[0015]

[0016] Where n is the amount of data, is the predicted value, y i The actual result of the test set.

[0017] Compared with the prior art, the present invention has found through research and analysis that the cumulative particle size distribution of nano-cerium dioxide and the transmittance of the cerium dioxide dispersion in the process of preparing the coating are key characteristics that affect the subsequent composite coating performance and coating performance (the cumulative particle size distribution characterizes the particle size distribution range and characteristics of the nanoparticles in the dispersion, and the transmittance is the transmittance of the organic solvent dispersion of nano-cerium dioxide, and its purpose is to evaluate the optical properties and dispersion uniformity of the dispersion). By using these two characteristics of cerium dioxide to characterize cerium dioxide for database construction, the complexity of the database is greatly reduced, and it is unnecessary to consider the influence of other preparation parameters in the preparation process of nano-cerium dioxide (such as the composition of the precursor solution, the synthesis temperature, the selection of the precipitant, the flushing solution, etc.) on the coating or coating performance, thereby simplifying the difficulty of model training and making the method of the present invention applicable to commercially available cerium dioxide without being limited to its preparation process.

[0018] The present invention further selects ultraviolet shielding performance and visible light transmittance as the key performance of the coating, which fully meets the application requirements of ultraviolet shielding coatings or coatings. Considering that after the components of the composite coating are determined, the heat treatment process has little effect on the coating performance, therefore, the heat treatment process is not used as an input feature, thereby further reducing the amount of data in the data set; finally, the cumulative particle size distribution of nano-cerium dioxide, organic solvents, types of fluorocarbon coatings, the ratio of fluorocarbon coatings to nano-cerium dioxide, and the transmittance of cerium dioxide dispersions and other characteristics closely related to performance are selected as input features.

[0019] The present invention adopts a variety of machine learning methods to construct a model, uses a database to train and cross-validate the machine learning model respectively, and obtains a trained machine learning model; uses a score function to evaluate the machine learning model, and selects the model with the best fitting effect. The obtained model can be used to predict the performance of nano-cerium dioxide ultraviolet shielding coatings. The method of the present invention does not require the configuration of composite coatings or the coating experiment. It only needs to be based on the characteristics of the composite coating to be configured (the cumulative particle size distribution of nano-cerium dioxide, surface modifiers, organic solvents, types of fluorocarbon coatings, the ratio of fluorocarbon coatings to nano-cerium dioxide, and the light transmittance of cerium dioxide dispersions) to obtain the performance prediction results of the composite coating and the coating after coating, thereby saving time and economic costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is the X-ray diffraction pattern of synthetic cerium dioxide;

[0021] Figure 2 This is a transmission electron microscope image of synthesized cerium dioxide;

[0022] Figure 3 is the UV-visible spectrum of synthetic cerium dioxide;

[0023] Figure 4 This is a comparison chart of the three models in terms of UV cutoff performance;

[0024] Figure 5 A summary diagram explaining the performance of the SHAP method on UV cutoff. DETAILED DESCRIPTION

[0025] The present invention is further described and illustrated below in conjunction with specific embodiments. The embodiments are merely exemplary of the present disclosure and do not define the scope of limitation. The technical features of each embodiment of the present invention may be combined accordingly without conflicting with each other.

[0026] The present invention provides a performance prediction method for nano-cerium dioxide ultraviolet shielding coatings based on machine learning optimization. The method prepares and tests the performance of the cerium dioxide ultraviolet shielding coatings by experimental means, obtains data, selects key characteristic parameters from the data, and constructs a database between the key characteristic parameters and the performance of the ultraviolet shielding coatings; a machine learning method is used to construct a machine learning model, the model is trained using the data in the database, and the best model is selected. When applied, for the nano-cerium dioxide ultraviolet shielding coatings that need to be predicted for performance, the cumulative particle size distribution of the nano-cerium dioxide, the surface modifier, the organic solvent, the type of fluorocarbon coating, the ratio of fluorocarbon coating to nano-cerium dioxide, and the light transmittance of the cerium dioxide dispersion are used as inputs and input into the selected model, so that the ultraviolet shielding performance and visible light transmittance of the ultraviolet shielding coating applied thereto can be predicted.

[0027] The specific process of the method of the present invention is as follows:

[0028] S1, preparing a precursor solution for preparing nano-cerium dioxide particles, adding a precipitant to synthesize nano-cerium dioxide precipitates by a precipitation method, centrifuging and washing the precipitates to obtain nano-cerium dioxide particles; changing the preparation conditions to prepare multiple groups of different nano-cerium dioxide particles;

[0029] S2, adding a surface modifier to each group of nano-cerium dioxide and dispersing it in an organic solvent to obtain nano-cerium dioxide particles with good dispersibility, mixing the organic solvent dispersion of nano-cerium dioxide with a fluorocarbon coating to obtain a composite coating, coating with the composite coating and preparing multiple groups of transparent coatings with nano-cerium dioxide through heat treatment, and testing the ultraviolet shielding performance and visible light transmittance of the coatings;

[0030] The surface modifier is one or more of phosphate, titanate and silane coupling agent. The phosphate is one or more of isooctyl phosphate, dodecyl phosphate, alkyl diphenyl phosphate, tetradecyl phosphate and lauryl alcohol polyoxyethylene ether. The titanate includes one or more of isopropyl tri(dodecylbenzenesulfonyl) titanate, bis(acetylacetonato) diisopropyl titanate, isopropyl tri(dioctyl pyrophosphate acyloxy) titanate, isopropyl tri(dioctyl phosphate acyloxy) titanate and isopropyl dioleyloxy(dioctyl phosphate acyloxy) titanate. The silane coupling agent is one or more of vinyl trimethoxy silane, vinyl triethoxy silane, dodecyl trimethoxy silane, dodecyl triethoxy silane, hexadecyl trimethoxy silane, hexadecyl triethoxy silane and phenyl trimethoxy silane.

[0031] The organic solvent is one or more of methanol, ethanol, isopropanol, n-butanol, tert-butanol, octanol, propylene glycol, benzyl alcohol, acetone, butanone, propylene glycol methyl ether acetate, tetrahydrofuran, dimethyl sulfoxide, chloroform, ethyl acetate, butyl acetate, benzene, toluene, xylene, heptane, n-hexane, cyclohexane and petroleum ether, and the fluorocarbon coating is one or more of polytetrafluoroethylene (PTFE), polyvinylidene fluoride (PVDF) and trifluorochloroethylene-vinyl ether copolymer (FEVE).

[0032] The heat treatment process has little effect on the performance of the obtained coating, and each coating used in the present invention adopts the same heat treatment process to prepare the coating. Typically but not limiting, the heat treatment is first high-temperature curing at a temperature of 100-200 degrees Celsius, and then low-temperature curing, and the low-temperature temperature is 40-80 degrees Celsius.

[0033] In the performance test of the present invention, the ultraviolet shielding performance test interval is 280-400 nanometers, characterized by ultraviolet cutoff rate; the visible light transmission test interval is 400-800 nanometers, characterized by visible light transmittance;

[0034]

[0035] Where λ is the wavelength and T(λ) is the transmittance at that wavelength.

[0036] S3, taking the cumulative particle size distribution of nano-cerium dioxide, the surface modifier in step S2, the organic solvent, the light transmittance of the dispersion, the type of fluorocarbon coating, and the ratio of fluorocarbon coating to nano-cerium dioxide as characteristic values, and the ultraviolet shielding performance and visible light transmittance of the coating as target variables, to establish a database;

[0037] S4, construct machine learning models based on different machine learning methods, and use the database of S3 to train and cross-validate the machine learning models to obtain trained machine learning models; use the scoring function to evaluate the machine learning models and select the model with the best fitting effect. The machine learning method is one or more of linear regression, support vector machine, random forest regression, ridge regression, and extreme gradient boosting method. The cross-validation uses k-fold cross-validation, which divides the data set into multiple subsets and uses different combinations of training sets and test sets to train and validate the model multiple times, thereby effectively evaluating the performance of the model on different data, reducing the risk of overfitting, and enhancing the generalization ability of the model. The scoring function is the determination coefficient R 2 The root mean square error (RMSE) measures the accuracy of the model's predictions, and the closer it is to 0, the smaller the prediction error.

[0038]

[0039] Where n is the amount of data, is the predicted value, y i The actual result of the test set.

[0040] S5. For the nano-cerium dioxide ultraviolet shielding coating that needs to be predicted for performance, the cumulative particle size distribution of the nano-cerium dioxide, the type of surface modifier, organic solvent, fluorocarbon coating, the ratio of fluorocarbon coating and nano-cerium dioxide, and the transmittance of the cerium dioxide dispersion are taken as input and input into the model selected in S4. This can predict the ultraviolet shielding performance and visible light transmittance of the ultraviolet shielding coating applied thereon.

[0041] In order to ensure the practicability of the model, the database of the present invention needs to contain different data of nano-cerium dioxide. Among them, a certain amount or cumulative particle size distribution of nano-cerium dioxide is obtained by commercial purchase, but the types of cerium dioxide that can be purchased on the market are usually limited, which is often insufficient to cover the range requirements of training data. Therefore, the present invention further provides a method for preparing nano-cerium dioxide to obtain more abundant nano-cerium dioxide. The preparation process of nano-cerium dioxide is: prepare a precursor solution for preparing nano-cerium dioxide particles, add a precipitant to synthesize nano-cerium dioxide precipitate by precipitation method, centrifuge and wash the precipitate to obtain nano-cerium dioxide particles.

[0042] Typically but not limiting, the precursor solution includes a cerium salt and a solvent, the cerium salt is one or more of cerium sulfate, cerium nitrate, cerium chloride, and cerium acetate, the solvent is one or more of water, methanol, ethanol, ethylene glycol, n-propanol, isopropanol, and propylene glycol, the precipitant is one or more of ammonia water, sodium hydroxide solution, potassium hydroxide solution, sodium bicarbonate solution, and sodium carbonate solution, the synthesis temperature is 40-90 degrees, the synthesis time is 2-20 hours, the centrifugal speed is 5000-11000 revolutions per minute, and the detergent is a solution of water and ethanol in different volume ratios, and the ratio of water to ethanol is 1:1-1:15.

[0043] The present invention uses the two characteristics of the cumulative particle size distribution of nano-cerium dioxide and the transmittance of the cerium dioxide dispersion in the process of preparing the coating to characterize cerium dioxide for constructing a database, eliminating the need to consider the effects of other preparation parameters in the preparation process of nano-cerium dioxide on the performance of the coating or coating, and also makes the method of the present invention applicable to commercially available cerium dioxide without being limited to its preparation process.

[0044] Example 1

[0045] Dissolve 21.7g of cerium nitrate in 800mL of a solution with a volume ratio of water to ethylene glycol of 1:3, stir for 0.5h, transfer to a reactor, heat to 60℃, slowly add 36mL of ammonia water, react for 5 hours, centrifuge after standing to precipitate, use a mixture of water and ethanol with a volume ratio of 1:2 as a detergent to wash the precipitate, the centrifugal speed is 10000rpm, after washing 3 times, cerium dioxide is obtained, its particle size is tested, and the cumulative particle size distribution is calculated. The synthesized cerium dioxide is characterized, such as Figure 1 is the X-ray diffraction pattern of synthetic cerium dioxide, corresponding to the PDF card of cerium dioxide, such as Figure 2 This is a transmission electron microscope image of synthetic cerium dioxide, with obvious lattice fringes and good crystallinity of the sample. Figure 3 This is the UV-visible spectrum of synthetic cerium dioxide, which has good absorption in the UV region and poor absorption in the visible light region.

[0046] Cerium dioxide and a surface modifier are dispersed in toluene, the surface modifier is isopropyl tri(dodecylbenzenesulfonyl) titanate, and the mass ratio of cerium dioxide: isopropyl tri(dodecylbenzenesulfonyl) titanate: toluene is 1:0.1:20. The transmittance is tested, and the toluene dispersion of cerium dioxide is mixed with fluorocarbon coating to obtain a composite coating. The coating is obtained by scraping, and is first dried at 100°C for 20 minutes and then dried at 60°C for 40 hours. The ultraviolet cutoff performance and visible light transmittance of the composite coating are tested.

[0047] Example 2

[0048] 26g of cerium nitrate was dissolved in 900mL of a solution with a volume ratio of water to propylene glycol of 1:2, stirred for 0.5h, transferred to a reactor, heated to 70°C, slowly added 40mL of ammonia water, reacted for 6 hours, centrifuged after standing and precipitating, and washed with a mixed solution with a volume ratio of water to ethanol of 1:2 as a detergent, the centrifugal speed was 10000rpm, and after washing 3 times, cerium dioxide was obtained, and the subsequent treatment steps were consistent with those in Example 1. Finally, the ultraviolet cutoff performance and visible light transmittance of the composite coating were tested.

[0049] Example 3

[0050] Cerium dioxide with different cumulative particle size distributions is obtained by commercially available or continuously changing synthesis conditions, and a cerium dioxide dispersion is obtained by selecting different surface modifiers and organic solvents, and the light transmittance of the dispersion is tested. Different types of fluorocarbon coatings and ratios of fluorocarbon coatings and nano-cerium dioxide are selected to configure composite coatings, and the coatings are coated, and the obtained coatings are tested for ultraviolet cutoff performance and visible light transmittance. The cumulative particle size distribution of nano-cerium dioxide, the surface modifier, organic solvent, light transmittance of the dispersion, the type of fluorocarbon coating, and the ratio of fluorocarbon coating to nano-cerium dioxide in step S2 are used as characteristic values, wherein the light transmittance is the light transmittance of the organic solvent dispersion of nano-cerium dioxide; the ultraviolet cutoff performance and visible light transmittance are used as target variables to establish a database.

[0051] Use ridge regression, random forest regression, and extreme gradient boosting method to build machine learning models respectively; use the data in the aforementioned database to train and k-fold cross-validate the three models to obtain three trained models.

[0052] According to the coefficient of determination R 2 The three models were evaluated with the value of root mean square error RMSE and the better model was selected, such as Figure 4 , we can get the performance of the three models. For the three models, the determination coefficient R of the model constructed by the extreme gradient boosting method is 2The maximum and the root mean square error RMSE is the smallest, so the model corresponding to the extreme gradient boosting method is selected, and then the SHAP analysis is performed to obtain the important features with greater influence, such as Figure 5 , a larger cumulative particle size distribution can be obtained, which has a greater impact on the results of UV cutoff performance and visible light transmittance performance.

[0053] For nano-cerium dioxide ultraviolet shielding materials that need to have their performance predicted, after determining their composition, there is no need to conduct coating compounding or coating experiments. The cumulative particle size distribution of the nano-cerium dioxide, surface modifiers, organic solvents, types of fluorocarbon coatings, the ratio of fluorocarbon coatings to nano-cerium dioxide, and the transmittance of the cerium dioxide dispersion that make up the materials can be directly input into the selected model, and the ultraviolet shielding performance and visible light transmittance of the ultraviolet shielding coating applied thereto can be predicted, thereby saving time and economic costs.

[0054] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. For ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A performance prediction method for nano-cerium dioxide ultraviolet shielding coating based on machine learning optimization, characterized in that: The steps include: S1, preparing a precursor solution for preparing nano-cerium dioxide particles, adding a precipitant to synthesize nano-cerium dioxide precipitates by a precipitation method, centrifuging and washing the precipitates to obtain nano-cerium dioxide particles; changing the preparation conditions to prepare multiple groups of different nano-cerium dioxide particles; S2, adding a surface modifier to each group of nano-cerium dioxide and dispersing it in an organic solvent to obtain nano-cerium dioxide particles with good dispersibility, mixing the organic solvent dispersion of nano-cerium dioxide with a fluorocarbon coating to obtain a composite coating, coating with the composite coating and preparing multiple groups of transparent coatings with nano-cerium dioxide through heat treatment, and testing the ultraviolet shielding performance and visible light transmittance of the coatings; S3, taking the cumulative particle size distribution of nano-cerium dioxide, the surface modifier in step S2, the organic solvent, the light transmittance of the dispersion, the type of fluorocarbon coating, and the ratio of fluorocarbon coating to nano-cerium dioxide as characteristic values, and the ultraviolet shielding performance and visible light transmittance of the coating as target variables, to establish a database; S4 builds machine learning models based on different machine learning methods, and uses the S3 database to train and cross-validate the machine learning models to obtain trained machine learning models; uses the scoring function to evaluate the machine learning models and selects the model with the best fitting effect. S5. For the nano-cerium dioxide ultraviolet shielding coating that needs to be predicted for performance, the cumulative particle size distribution of the nano-cerium dioxide, the type of surface modifier, organic solvent, fluorocarbon coating, the ratio of fluorocarbon coating and nano-cerium dioxide, and the transmittance of the cerium dioxide dispersion are taken as input and input into the model selected in S4. This can predict the ultraviolet shielding performance and visible light transmittance of the ultraviolet shielding coating applied thereon.

2. The performance prediction method according to claim 1, characterized in that: The precursor solution includes a cerium salt and a solvent, wherein the cerium salt is one or more of cerium sulfate, cerium nitrate, cerium chloride, and cerium acetate, and the solvent is one or more of water, methanol, ethanol, ethylene glycol, n-propanol, isopropanol, and propylene glycol. The precipitant is one or more of ammonia water, sodium hydroxide solution, potassium hydroxide solution, sodium bicarbonate solution, and sodium carbonate solution. The synthesis temperature is 40-90 degrees, the synthesis time is 2-20 hours, the centrifugal speed is 5000-11000 revolutions per minute, and the detergent is a solution of water and ethanol in different volume ratios, and the ratio of water to ethanol is 1:1-1:

15.

3. The performance prediction method according to claim 1, characterized in that: The surface modifier described in S2 is one or more of phosphate, titanate and silane coupling agent, the phosphate is one or more of isooctyl phosphate, dodecyl phosphate, alkyl diphenyl phosphate, tetradecyl phosphate and lauryl alcohol polyoxyethylene ether, the titanate includes one or more of isopropyl tri(dodecylbenzenesulfonyl) titanate, bis(acetylacetonato) diisopropyl titanate, isopropyl tri(dioctyl pyrophosphate acyloxy) titanate, isopropyl tri(dioctyl phosphate acyloxy) titanate and isopropyl dioleyloxy(dioctyl phosphate acyloxy) titanate, the silane coupling agent is one or more of vinyl trimethoxy silane, vinyl triethoxy silane, dodecyl trimethoxy silane, dodecyl triethoxy silane, hexadecyl trimethoxy silane, hexadecyl triethoxy silane and phenyl trimethoxy silane.

4. The performance prediction method according to claim 1, characterized in that: The organic solvent described in S2 is one or more of methanol, ethanol, isopropanol, n-butanol, tert-butanol, octanol, propylene glycol, benzyl alcohol, acetone, butanone, propylene glycol methyl ether acetate, tetrahydrofuran, dimethyl sulfoxide, chloroform, ethyl acetate, butyl acetate, benzene, toluene, xylene, heptane, n-hexane, cyclohexane and petroleum ether, and the fluorocarbon coating is one or more of polytetrafluoroethylene (PTFE), polyvinylidene fluoride (PVDF), and trifluorochloroethylene-vinyl ether copolymer (FEVE).

5. The performance prediction method according to claim 1, characterized in that: The heat treatment in S2 is first high-temperature curing at a temperature of 100-200 degrees Celsius, and then low-temperature curing, wherein the low-temperature temperature is 40-80 degrees Celsius.

6. The performance prediction method according to claim 1, characterized in that: In S4, the machine learning method is one or more of linear regression, support vector machine, random forest regression, ridge regression, and extreme gradient boosting method.

7. The performance prediction method according to claim 1, characterized in that: The cross-validation uses k-fold cross-validation, which divides the data set into multiple subsets and uses different combinations of training sets and test sets to train and validate the model multiple times, thereby effectively evaluating the performance of the model on different data, reducing the risk of overfitting, and enhancing the generalization ability of the model.

8. The performance prediction method according to claim 1, characterized in that: The score function is the coefficient of determination R 2 The root mean square error (RMSE) measures the accuracy of the model's predictions, and the closer it is to 0, the smaller the prediction error. Where n is the amount of data, is the predicted value, y i The actual result of the test set.

9. The performance prediction method according to claim 1, characterized in that: The UV shielding performance test interval is 280-400 nanometers, characterized by UV cutoff rate; the visible light transmittance test interval is 400-800 nanometers, characterized by visible light transmittance; Where λ is the wavelength and T(λ) is the transmittance at that wavelength.

Images