Photocuring material performance prediction method

Abstract

The invention belongs to the technical field of 3D printing, and particularly relates to a photocuring material performance prediction method, which comprises the steps of collecting a first information set of photocuring material formula data, obtaining a second information set of illumination condition data corresponding to a material formula, and collecting a third information set of 3D printing process parameters set for illumination conditions. And preparing a training sample by using the three information sets, and designing and training a multi-layer neural network structure to predict the performance of the photocuring material. And by evaluating the prediction result, adjusting the printing parameters to form an optimized parameter set, and applying the optimized parameters to new performance prediction to verify the improvement effect. And finally, adjusting the neural network according to actual performance data feedback to improve prediction precision. Through the efficient neural network model, the experimental dependence is reduced, the research and development period is shortened, the cost is reduced, and the 3D printing product quality and the production efficiency are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of 3D printing, and in particular relates to a method for predicting properties of photocurable materials. Background Art

[0002] Traditional methods for evaluating the performance of photocurable materials usually rely on a large number of experimental tests, which is not only time-consuming but also costly. With the development of 3D printing technology, higher requirements are placed on quickly and accurately predicting the performance of photocurable materials under different conditions. At present, most research focuses on how to improve material formulations or optimize printing parameters through experimental means, but this method is inefficient and difficult to adapt to rapidly changing market demands.

[0003] Existing methods for evaluating the performance of photocurable materials rely too much on experimental verification, resulting in a long R&D cycle and high costs for new materials, and it is difficult to adjust printing parameters in real time to cope with different lighting conditions and changes in material formulations. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting the properties of photocurable materials. By constructing a multi-level neural network model, it is possible to efficiently and accurately predict the mechanical properties, thermal stability and optical properties of the material when the material formula, lighting conditions and printing parameters are input, so as to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention adopts the following technical solution: a method for predicting properties of a photocurable material, comprising the following steps:

[0006] S1: Collecting a first information set including photocurable material recipe data;

[0007] S2: Based on the first information set, obtaining a second information set corresponding to the illumination condition data, wherein the second information set is directly associated with the material formula in the first information set;

[0008] S3: Based on the second information set, collecting a third information set related to 3D printing process parameters, and ensuring that these parameters are set for the lighting conditions in the second information set;

[0009] S4: preparing training samples using the data obtained from the first to third information sets, each sample covering a recipe, a lighting condition, and a printing parameter;

[0010] S5: Based on the prepared training samples, design and train a multi-layer neural network structure to predict the performance of light-curable materials. Use the trained neural network structure to input new formula data to evaluate its performance under specific lighting conditions.

[0011] S6: Based on the evaluation results, the printing parameters are adjusted to form an optimized printing parameter set. The optimized printing parameters are applied to the new light-curing material performance prediction to verify the actual performance of the material under the improved printing conditions.

[0012] S7: Based on the actual performance data, feedback is provided to adjust the multi-layer neural network to improve prediction accuracy, and the iteration is continued until the expected prediction accuracy is achieved.

[0013] Preferably, in S1, the following sub-steps are included:

[0014] Quantify the proportion of each component in the first information set, using linear combination where x i represents the proportion of the i-th material, w i is the corresponding weight coefficient;

[0015] Using the quantified F(x) value, calculate the corresponding material mixture density D, through the formula where d i is the known density of the ith material;

[0016] Based on the calculated density D, the basic physical properties of the material mixture are evaluated.

[0017] Preferably, in S2, the following sub-steps are included:

[0018] For each light-curable material formula in the first information set, record its light intensity I and light exposure time t under a specific light source to form a second information set, where light intensity and time are parameters that directly affect material curing;

[0019] Based on the recorded illumination conditions, calculate the photopolymerization rate constant, k, using the formula Where T represents the transmittance under given lighting conditions;

[0020] The rate constant k is used to predict the curing degree C of the material under different lighting conditions. The curing degree is calculated by the formula C = 1-e -k·I·t to be sure.

[0021] Preferably, in S3, the following sub-steps are included:

[0022] For each illumination condition in the second information set, the parameters of nozzle speed v, layer thickness h and filling density p during the 3D printing process are recorded to form a third information set;

[0023] Calculate the single-layer printing time T based on the nozzle speed v and layer thickness h l , using the formula Where l is the layer length, t is the time the nozzle waits before printing a new layer. lMatch the curing time under lighting conditions;

[0024] Using the obtained single-layer printing time T l , combined with the filling density p, the actual curing quality Q of each layer of material under given printing parameters is evaluated, and the formula Calculate, where t cure is the curing time determined based on the second information set.

[0025] Preferably, in S4, the following sub-steps are included:

[0026] Integrate the material formula information in the first information set, the lighting condition data in the second information set, and the 3D printing parameters in the third information set to form a structured data entry;

[0027] According to the data entry, define the feature vector X of each training sample, where X = [x1, x2, x3, ..., x n ], x1 represents the material formula, x2 represents the lighting conditions, and x3 represents the printing parameters;

[0028] For each feature vector X, a corresponding performance label Y is assigned, which is based on the material properties obtained from experimental measurements to form a training sample pair (X, Y) for subsequent analysis.

[0029] Preferably, in S5, the following sub-steps are included:

[0030] The prepared training sample pair (X, Y) is input into a pre-designed multi-layer neural network, where the input layer receives the feature vector X and the output layer produces the performance prediction

[0031] Using loss function To quantify the predicted value The difference between the actual label Y and the loss function is defined as Represents the squared loss of prediction error;

[0032] According to the calculated loss, the weights W of each layer in the network are updated through the back-propagation algorithm to minimize the loss function, and the optimization is iterated until the network can accurately predict the performance of the photocurable material.

[0033] Preferably, in S5, the following sub-steps are included:

[0034] Create feature vector X for new recipe data new , where X new =[x 1,new ,x 2,new ,x 3,new ,...,x n,new ], x 1,newIndicates the information of the new recipe, x 2,new represents the corresponding lighting conditions, x 3,new Indicates the printing parameters that match it;

[0035] The constructed feature vector X new Input into the trained neural network to obtain the predicted performance output in reflects the expected performance of new materials under specific lighting conditions;

[0036] based on Compare the goodness of fit R between the predicted performance and the actual test results using the formula where Y actual is the performance value obtained from actual testing. Yes actual The average value of .

[0037] Preferably, in S6, the following sub-steps are included:

[0038] Analyze the performance difference ΔP in the evaluation results, Identify the printing parameters that have the greatest impact on performance;

[0039] According to the most influential parameter p max , apply the fine-tuning formula p′ max =p max +Δp, where Δp is the increment adjusted based on the performance difference ΔP to optimize the printing parameters;

[0040] Using the optimized parameter p′ max , reconfigure the printing parameter set P′, and predict the performance again through the neural network.

[0041] Preferably, in S6, the following sub-steps are included:

[0042] Use the optimized printing parameter set P′ to construct new training samples (X′, Y′), where X′ contains the adjusted printing parameters and other related variables;

[0043] Input the new training sample X′ into the previously trained multi-layer neural network to obtain the performance prediction value under the improved printing condition

[0044] Compare predicted values Compared with the actual test result Y′, the performance improvement index I is calculated using the formula Where Y is the performance value measured under original conditions.

[0045] Preferably, in S7, the following sub-steps are included:

[0046] The actual performance data Y″ and the corresponding predicted values ​​collected A new training sample pair (X″, Y″) is formed, where Y″ refers to the performance data of the photocurable material obtained through actual testing or experiments after using the optimized printing parameter set P′. It refers to the results obtained by predicting the performance of new light-curing materials using a trained multi-layer neural network under the same conditions;

[0047] Using the new training sample pair (X″, Y″), the weight W of the multi-layer neural network is updated by the gradient descent method. The update rule is: Where η is the learning rate, is the gradient of the loss function with respect to the weights;

[0048] According to the updated weight W new , re-evaluate the prediction accuracy A of the neural network, using the formula in is the new predicted value after updating the weights, The average value of is calculated and the iteration is continued until the prediction accuracy meets the predetermined standard.

[0049] Technical effects and advantages of the present invention: Compared with the prior art, the method for predicting the performance of a photocurable material proposed by the present invention has the following advantages:

[0050] The present invention constructs a multi-level neural network model, which can efficiently and accurately predict the mechanical properties, thermal stability and optical properties of materials by inputting material formula, lighting conditions and printing parameters. This method reduces the dependence on traditional experimental methods, greatly shortens the cycle of new material research and development, reduces research and development costs, and can quickly adjust printing parameters according to actual needs, thereby significantly improving the quality and production efficiency of 3D printed products. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is a flow chart of the method for predicting properties of photocurable materials of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0053] The present invention provides a method for predicting the performance of photocurable materials. By constructing a multi-level neural network model, the mechanical properties, thermal stability and optical properties of the material can be predicted efficiently and accurately by inputting the material formula, lighting conditions and printing parameters. This method reduces the reliance on traditional experimental methods, greatly shortens the research and development cycle of new materials, reduces research and development costs, and can quickly adjust printing parameters according to actual needs, thereby significantly improving the quality and production efficiency of 3D printed products. The details are as follows:

[0054] like Figure 1 As shown, a method for predicting properties of a photocurable material in this embodiment includes the following steps:

[0055] S1: Collecting a first information set including photocurable material recipe data; specifically as follows:

[0056] Quantify the proportion of each component in the first information set, using linear combination where x i represents the proportion of the i-th material, w i is the corresponding weight coefficient; using the quantified F(x) value, calculate the corresponding material mixture density D, through the formula where d i is the known density of the ith material;

[0057] Based on the calculated density D, the basic physical properties of the material mixture are evaluated. Making the evaluation of the material formula more scientific and accurate helps to screen out material combinations with ideal properties at an early stage, thereby reducing the number of subsequent experiments, speeding up the research and development process, and ensuring that the final selected material formula can show excellent performance in practical applications. In addition, the material density information obtained in this way can also provide an important reference for the optimization of lighting conditions and printing parameters in subsequent steps, further improving the consistency and reliability of 3D printed products.

[0058] S2: Based on the first information set, a second information set corresponding to the illumination condition data is obtained, wherein the second information set is directly associated with the material formula in the first information set; specifically as follows:

[0059] For each light-curable material formula in the first information set, record its light intensity I and light exposure time t under a specific light source to form a second information set, where light intensity and time are parameters that directly affect material curing;

[0060] Based on the recorded illumination conditions, calculate the photopolymerization rate constant, k, using the formula Wherein T represents the transmittance under given lighting conditions; the transmittance T can be obtained through experimental measurement or by querying an existing material property database.

[0061] The rate constant k is used to predict the curing degree C of the material under different lighting conditions. The curing degree is calculated by the formula C = 1-e -k·I·t This formula reflects the degree of curing of the material under specific lighting conditions, that is, the degree of curing reaction of the material under the action of light.

[0062] By recording the light intensity and time of each formula under a specific light source, a set of light condition data directly related to the material formula can be formed. Then, the photopolymerization reaction rate constant calculated based on this data allows us to predict the degree of curing of the material under different light conditions. This not only helps to understand the material curing mechanism, but also provides an important basis for the subsequent optimization of printing parameters.

[0063] By calculating the photopolymerization reaction rate constant and the degree of curing, the curing effect of the material under specific lighting conditions can be predicted more accurately, thereby improving the accuracy of the overall prediction model. Based on the predicted results of the degree of curing, the lighting conditions in the 3D printing process can be set more reasonably to ensure that the performance of the material during the curing process meets expectations. Through precise control of the curing process, the material performance can be optimized and the quality and reliability of the final product can be improved.

[0064] S3: Based on the second information set, a third information set related to 3D printing process parameters is collected to ensure that these parameters are set for the lighting conditions in the second information set; specifically, as follows:

[0065] For each lighting condition in the second information set, the parameters of nozzle speed v, layer thickness h and filling density p during the 3D printing process are recorded to form a third information set; these parameters directly affect the curing effect of the material during the printing process and the quality of the final product.

[0066] Calculate the single-layer printing time T based on the nozzle speed v and layer thickness h l , using the formula Where l is the layer length, t is the time the nozzle waits before printing a new layer. l Match the curing time under lighting conditions; the purpose of this is to ensure that the single-layer printing time matches the curing time under lighting conditions to ensure that each layer of material can be fully cured.

[0067] Using the obtained single-layer printing time T l , combined with the filling density p, the actual curing quality Q of each layer of material under given printing parameters is evaluated, and the formula Calculate, where t cure The curing time is determined based on the second information set. This formula can quantify the degree of curing of each layer of material under specific printing parameters to ensure that the material achieves the expected curing effect during the printing process.

[0068] By calculating the single-layer printing time and curing quality, it is possible to ensure that each layer of material reaches the optimal state during the curing process, avoiding product defects caused by insufficient or excessive curing. Based on the evaluation results of the curing quality, parameters such as nozzle speed, layer thickness, and filling density can be adjusted in a targeted manner to make the 3D printing process more efficient and stable, thereby improving the quality of the final product. By ensuring the curing quality of each layer of material, the overall consistency and reliability of 3D printed products can be significantly improved, the scrap rate can be reduced, and production efficiency can be improved. Based on the results of the curing quality evaluation, it can be fed back to the previous material formula and lighting condition settings to form a closed-loop optimization system, and continuously improve the scientificity and rationality of the entire process.

[0069] S4: Using the data obtained from the first to third information sets, prepare training samples, each sample covering a recipe, a lighting condition, and a printing parameter; specifically, as follows:

[0070] The material formula information in the first information set, the lighting condition data in the second information set, and the 3D printing parameters in the third information set are integrated to form structured data entries; these entries contain a complete description of the experimental conditions, which facilitates subsequent data processing and analysis.

[0071] According to the data entry, define the feature vector X of each training sample, where X = [x1, x2, x3, ..., x n ], x1 represents the material formula, x2 represents the lighting conditions, and x3 represents the printing parameters; each eigenvector X is a multidimensional vector containing all influencing factors and can fully describe the conditions of an experimental sample.

[0072] For each feature vector X, a corresponding performance label Y is assigned. The label is based on the material properties measured experimentally, such as mechanical strength, thermal stability, optical properties, etc., to form a training sample pair (X, Y) for subsequent analysis, where X represents the input feature and Y represents the output label. These training sample pairs will be used for subsequent model training and performance prediction.

[0073] By integrating multiple information sets into structured data entries, data consistency and comparability are ensured, laying a solid foundation for subsequent model training. Material formulas, lighting conditions, and printing parameters are converted into feature vectors, allowing machine learning models to effectively process these multivariate data, thereby improving the accuracy of predictions. By assigning performance labels based on experimental measurements to each feature vector, it can be directly used to train multi-layer neural networks, allowing the model to learn the complex relationship between material properties and input conditions. Standardized training sample pairs can improve the learning efficiency of the model, allowing the prediction model to better generalize to new experimental conditions, thereby providing reliable performance predictions in practical applications.

[0074] S5: Based on the prepared training samples, design and train a multi-layer neural network structure to predict the performance of light-curable materials. Use the trained neural network structure to input new formula data to evaluate its performance under specific lighting conditions; the details are as follows:

[0075] The prepared training sample pair (X, Y) is input into a pre-designed multi-layer neural network, where the input layer receives the feature vector X and the output layer produces the performance prediction That is, the predicted material performance value; using the loss function To quantify the predicted value The difference between the actual label Y and the loss function is defined as Represents the square loss of the prediction error; based on the calculated loss, the weights W of each layer in the network are updated through the back propagation algorithm to minimize the loss function, and the optimization is iterated until the network can accurately predict the performance of the photocurable material.

[0076] Create feature vector X for new recipe data new , where X new =[x 1,new ,x 2,new ,x 3,new ,...,x n,new ], x 1,new Indicates the information of the new recipe, x 2,new represents the corresponding lighting conditions, x 3,new Represents the printing parameters that match it; the constructed feature vector X new Input into the trained neural network to obtain the predicted performance output in reflects the expected performance of new materials under specific lighting conditions;

[0077] based on Compare the goodness of fit R between the predicted performance and the actual test results using the formula where Y actual is the performance value obtained from actual testing. Yes actual The average value of .

[0078] By training a multi-layer neural network, it is possible to learn the complex relationship between material properties and formulations, lighting conditions and printing parameters, thereby improving the accuracy of predictions. The neural network model can capture nonlinear relationships, making the prediction results closer to reality.

[0079] By using trained neural networks, the performance of new materials under specific conditions can be quickly evaluated, reducing the number of experimental trial and error and speeding up the material development cycle. By evaluating the consistency between the predicted performance and the actual test results, the reliability of the prediction model can be ensured, thereby improving the performance consistency of the final product. By reducing the number of unnecessary experimental verifications, the cost of material development is effectively reduced and economic benefits are improved.

[0080] S6: Based on the evaluation results, the printing parameters are adjusted to form an optimized printing parameter set. The optimized printing parameters are applied to the new light-curing material performance prediction to verify the actual performance of the material under the improved printing conditions; the details are as follows:

[0081] Analyze the performance difference ΔP in the evaluation results, Determine the part of the printing parameters that has the greatest impact on performance; based on the identified parameters with the greatest impact p max , apply the fine-tuning formula p′ max =p max +Δp, where Δp is the increment adjusted based on the performance difference ΔP to optimize the printing parameters; using the optimized parameters p′ max , reconfigure the printing parameter set P′, and predict the performance again through the neural network.

[0082] Use the optimized printing parameter set P′ to construct a new training sample (X′, Y′), where X′ contains the adjusted printing parameters and other relevant variables; input the new training sample X′ into the previously trained multi-layer neural network to obtain the performance prediction value under the improved printing condition

[0083] Compare predicted values Compared with the actual test result Y′, the performance improvement index I is calculated using the formula Where Y is the performance value measured under original conditions. The performance improvement index II is used to evaluate whether the optimized printing parameters significantly improve the performance of the material.

[0084] By analyzing the performance difference ΔP, the main printing parameters that affect the material performance are identified and adjusted to optimize the printing parameter set. This helps to improve the performance of the material under specific printing conditions. By reconstructing the training samples and making predictions using the optimized printing parameter set, the performance of the model in actual applications can be verified to ensure that the model can accurately reflect the optimized material performance.

[0085] The optimized printing parameters can ensure that the material can achieve the expected performance level under different printing conditions, thereby improving the quality and reliability of the final product. By continuously adjusting the printing parameters based on the evaluation results, the optimal printing conditions can be quickly found, the number of unnecessary experiments can be reduced, and the speed of new product development can be accelerated.

[0086] S7: Based on the actual performance data, feedback is given to adjust the multi-layer neural network to improve the prediction accuracy, and it is continuously iterated until the expected prediction accuracy is achieved; the details are as follows:

[0087] The actual performance data Y″ and the corresponding predicted values ​​collected A new training sample pair (X″, Y″) is formed, where Y″ refers to the performance data of the photocurable material obtained through actual testing or experiments after using the optimized printing parameter set P′. It refers to the results obtained by predicting the performance of new light-curing materials using a trained multi-layer neural network under the same conditions;

[0088] Using the new training sample pair (X″, Y″), the weight W of the multi-layer neural network is updated by the gradient descent method. The update rule is: Where η is the learning rate, is the gradient of the loss function with respect to the weights;

[0089] According to the updated weight W new , re-evaluate the prediction accuracy A of the neural network, using the formula in is the new predicted value after updating the weights, The average value of is calculated and the iteration is continued until the prediction accuracy meets the predetermined standard.

[0090] By collecting actual performance data and predicted values, new training sample pairs are formed, and these sample pairs are used to update the weights of the neural network, so that the model can better fit the actual data, thereby improving the accuracy of the prediction. Through continuous iteration and weight update, the neural network can gradually approach the optimal state, so that the prediction accuracy continues to improve until the expected accuracy standard is reached. With the continuous increase of actual performance data, the model can learn more practical application scenarios, improve its generalization ability under different conditions, and enhance the robustness of the model.

[0091] In summary, the present invention collects a first information set of photocurable material formula data, obtains a second information set of illumination condition data corresponding to the material formula, and collects a third information set of 3D printing process parameters set for the illumination conditions. These three information sets are used to prepare training samples, and a multi-layer neural network structure is designed and trained to predict the performance of photocurable materials. By evaluating the prediction results, the printing parameters are adjusted to form an optimized parameter set, and the optimized parameters are applied to the new performance prediction to verify the improvement effect. Finally, the neural network is adjusted according to the feedback of the actual performance data to improve the prediction accuracy. The present invention reduces experimental dependence, shortens the R&D cycle, reduces costs, and improves the quality and production efficiency of 3D printed products through an efficient neural network model.

[0092] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting properties of photocurable materials, characterized in that: The following steps are involved: S1: Collecting a first information set including photocurable material recipe data; S2: Based on the first information set, obtaining a second information set corresponding to the illumination condition data, wherein the second information set is directly associated with the material formula in the first information set; S3: Based on the second information set, collecting a third information set related to 3D printing process parameters, and ensuring that these parameters are set for the lighting conditions in the second information set; S4: preparing training samples using the data obtained from the first to third information sets, each sample covering a recipe, a lighting condition, and a printing parameter; S5: Based on the prepared training samples, design and train a multi-layer neural network structure to predict the performance of light-curable materials. Use the trained neural network structure to input new formula data to evaluate its performance under specific lighting conditions. S6: Based on the evaluation results, the printing parameters are adjusted to form an optimized printing parameter set. The optimized printing parameters are applied to the new light-curing material performance prediction to verify the actual performance of the material under the improved printing conditions. S7: Based on the actual performance data, feedback is provided to adjust the multi-layer neural network to improve prediction accuracy, and the iteration is continued until the expected prediction accuracy is achieved.

2. A method for predicting properties of photocurable materials according to claim 1, characterized in that: In S1, the following sub-steps are included: Quantify the proportion of each component in the first information set, using linear combination where x i represents the proportion of the i-th material, w i is the corresponding weight coefficient; Using the quantified F(x) value, calculate the corresponding material mixture density D, through the formula where d i is the known density of the ith material; Based on the calculated density D, the basic physical properties of the material mixture are evaluated.

3. A method for predicting properties of photocurable materials according to claim 2, characterized in that: In S2, the following sub-steps are included: For each light-curable material formula in the first information set, record its light intensity I and light exposure time t under a specific light source to form a second information set, where light intensity and time are parameters that directly affect material curing; Based on the recorded illumination conditions, calculate the photopolymerization rate constant, k, using the formula Where T represents the transmittance under given lighting conditions; The rate constant k is used to predict the curing degree C of the material under different lighting conditions. The curing degree is calculated by the formula C = 1-e -k·I·t to be sure.

4. A method for predicting properties of photocurable materials according to claim 3, characterized in that: In S3, the following sub-steps are included: For each illumination condition in the second information set, the parameters of nozzle speed v, layer thickness h and filling density p during the 3D printing process are recorded to form a third information set; Calculate the single-layer printing time T based on the nozzle speed v and layer thickness h l , using the formula Where l is the layer length, t is the time the nozzle waits before printing a new layer. l Match the curing time under lighting conditions; Using the obtained single-layer printing time T l , combined with the filling density p, the actual curing quality Q of each layer of material under given printing parameters is evaluated, and the formula Calculate, where t cure is the curing time determined based on the second information set.

5. A method for predicting properties of photocurable materials according to claim 4, characterized in that: In S4, the following sub-steps are included: Integrate the material formula information in the first information set, the lighting condition data in the second information set, and the 3D printing parameters in the third information set to form a structured data entry; According to the data entry, define the feature vector X of each training sample, where X = [x1, x2, x3, ..., x n ], x1 represents the material formula, x2 represents the lighting conditions, and x3 represents the printing parameters; For each feature vector X, a corresponding performance label Y is assigned, which is based on the material properties obtained from experimental measurements to form a training sample pair (X, Y) for subsequent analysis.

6. A method for predicting properties of photocurable materials according to claim 5, characterized in that: In S5, the following sub-steps are included: The prepared training sample pair (X, Y) is input into a pre-designed multi-layer neural network, where the input layer receives the feature vector X and the output layer produces the performance prediction Using loss function To quantify the predicted value The difference between the actual label Y and the loss function is defined as represents the squared loss of prediction error; According to the calculated loss, the weights W of each layer in the network are updated through the back-propagation algorithm to minimize the loss function, and the optimization is iterated until the network can accurately predict the performance of the photocurable material.

7. A method for predicting properties of photocurable materials according to claim 6, characterized in that: In S5, the following sub-steps are included: Create feature vector X for new recipe data new , where X new =[x 1,new ,x 2,new ,x 3,new ,...,x n,new ], x 1,new Indicates the information of the new recipe, x 2,new represents the corresponding lighting conditions, x 3,new Indicates the printing parameters that match it; The constructed feature vector X new Input into the trained neural network to obtain the predicted performance output in reflects the expected performance of new materials under specific lighting conditions; based on Compare the goodness of fit R between the predicted performance and the actual test results using the formula where Y actual is the performance value obtained from actual testing. Yes actual The average value of .

8. A method for predicting properties of photocurable materials according to claim 7, characterized in that: In S6, the following sub-steps are included: Analyze the performance difference ΔP in the evaluation results, Identify the printing parameters that have the greatest impact on performance; According to the most influential parameter p max , apply the fine-tuning formula p′ max =p max +Δp, where Δp is the increment adjusted based on the performance difference ΔP to optimize the printing parameters; Using the optimized parameter p′ max , reconfigure the printing parameter set P′, and predict the performance again through the neural network.

9. A method for predicting properties of photocurable materials according to claim 8, characterized in that: In S6, the following sub-steps are included: Use the optimized printing parameter set P′ to construct new training samples (X′, Y′), where X′ contains the adjusted printing parameters and other related variables; Input the new training sample X′ into the previously trained multi-layer neural network to obtain the performance prediction value under the improved printing condition Compare predicted values Compared with the actual test result Y′, the performance improvement index I is calculated using the formula Where Y is the performance value measured under original conditions.

10. A method for predicting properties of photocurable materials according to claim 9, characterized in that: In S7, the following sub-steps are included: The actual performance data Y″ and the corresponding predicted values ​​collected A new training sample pair (X″, Y″) is formed, where Y″ refers to the performance data of the photocurable material obtained through actual testing or experiments after using the optimized printing parameter set P′. It refers to the results obtained by predicting the performance of new light-curing materials using a trained multi-layer neural network under the same conditions; Using the new training sample pair (X″, Y″), the weight W of the multi-layer neural network is updated by the gradient descent method. The update rule is: Where η is the learning rate, is the gradient of the loss function with respect to the weights; According to the updated weight W new , re-evaluate the prediction accuracy A of the neural network, using the formula in is the new predicted value after updating the weights, The average value of is calculated and the iteration is continued until the prediction accuracy meets the predetermined standard.

Images