Method for auxiliary development of MOFs modified polyurethane material based on machine learning

Abstract

The invention relates to the technical field of computational chemistry and foaming materials, and discloses a method for auxiliary development of an MOFs modified polyurethane material based on machine learning. The method comprises the following steps: S1, selecting a plurality of MOFs modified polyurethane materials, and determining a limit oxygen index through an experiment; determining the characteristic attribute value of the MOFs modified polyurethane material; s2, constructing a database by using the limit oxygen index and the determined characteristic attribute value; s3, dividing the database into a test set and a training set, and training by adopting various machine learning models to obtain a function relationship between the limit oxygen index and the feature attribute value; and S4, predicting limit oxygen indexes of various MOFs modified polyurethane materials according to the function relationship, and screening out the MOFs modified polyurethane material with the highest limit oxygen index. According to the method, material development is carried out in combination with machine learning, the economic and time cost brought by a large number of experiment trials and errors is saved, and the limit oxygen index of the screened MOFs modified polyurethane material is high.

Description

Technical Field

[0001] The present invention relates to the technical fields of computational chemistry and foaming materials, and particularly relates to a method for developing MOF-modified polyurethane materials assisted by machine learning. Background Art

[0002] Polyurethane has the advantages of light weight, heat insulation, sound insulation, shock absorption, low price, etc., and has important uses in automotive interior parts and household appliances. However, the disadvantage of this material is its poor flame retardancy, fast burning speed and excessive dripping during the burning process, which easily leads to the accelerated spread of fire. Therefore, it is very necessary to modify polyurethane materials to improve their flame retardancy. Metal-organic framework materials (MOFs) are porous materials formed by the bridging self-assembly of inorganic metal centers and organic ligands, and have the characteristics of large specific surface area, good thermal stability and structural stability, and containing catalytic components. These characteristics meet the basic requirements for them to be used as flame retardants. However, there are a wide variety of metal centers and organic ligands in MOFs, and the structure, pore size and functionality of MOFs can be regulated by selecting different metal ions and organic ligands, which has led to thousands of types of MOFs. At present, it is still unclear which MOF materials modified polyurethane have excellent flame-retardant performance. Repeated trial and error through experimental methods will consume a large amount of manpower and material resources. Therefore, it is of great significance to develop a method for quickly screening and predicting the flame retardant performance of MOF-modified polyurethane materials. Summary of the Invention

[0003] The purpose of the present invention is to overcome the problem in the prior art that it is necessary to repeatedly try and error through a large number of experiments to select MOF-modified polyurethane materials with excellent flame-retardant performance, and to provide a method for developing MOF-modified polyurethane materials assisted by machine learning. This method combines machine learning for the development of MOF-modified polyurethane materials, saving the economic and time costs brought by a large number of experimental trials and errors.

[0004] To achieve the above purpose, the present invention provides a method for developing MOF-modified polyurethane materials assisted by machine learning, which includes the following steps:

[0005] S1. Randomly select a variety of MOF-modified polyurethane materials, and determine their limiting oxygen index through experiments; determine the characteristic attribute values of the selected MOF-modified polyurethane materials;

[0006] S2. Construct a database with the limiting oxygen index and the determined characteristic attribute values;

[0007] S3. Divide the database into a test set and a training set, and use a variety of machine learning models for training to obtain the functional relationship between the limiting oxygen index and the characteristic attribute values;

[0008] S4. Predict the limiting oxygen index of various MOF-modified polyurethane materials according to the functional relationship, and screen out the MOF-modified polyurethane material with the highest limiting oxygen index.

[0009] Preferably, in step S1, the number of randomly selected MOF-modified polyurethane materials can be 80 - 120 kinds.

[0010] Preferably, the preparation process of the MOF-modified polyurethane material includes: under the presence of a mixed gas containing argon and oxygen, performing plasma treatment on the MOF material, then dispersing the plasma-treated MOF material in an organic solvent for ultrasonic treatment to obtain solution A, then dissolving polyurethane in the organic solvent to obtain solution B, then mixing solution A and solution B, and finally freeze-drying the mixed solution to obtain the MOF-modified polyurethane material.

[0011] Preferably, in step S1, the process of determining the characteristic attribute values of the selected MOF-modified polyurethane materials includes:

[0012] Select N characteristic attribute values of the MOF-modified polyurethane material, and calculate the pearson correlation coefficient between any two characteristic attribute values, with a total calculation of times;

[0013] When the pearson correlation coefficient of two characteristic attribute values > 0.9, then select one of the two characteristic attribute values and put it into the database;

[0014] When the pearson correlation coefficient of two characteristic attribute values ≤ 0.9, then both of the two characteristic attribute values are put into the database.

[0015] Preferably, in step S1, the characteristic attribute values of the selected MOF-modified polyurethane materials include the molecular weight X1 of the organic ligand, the number of oxygen atoms X2 of the organic ligand, the number of hydrogen atoms X3 of the organic ligand, the number of hydrogen atoms X4 of the organic ligand, the number of nitrogen atoms X5 of the organic ligand, the relative atomic mass X6 of the central metal, the atomic number X6 of the central metal, the number of electrons in the d orbital X7 of the central metal, the coordination number X8 of the central metal, the main group number X9 of the central metal element, the main group number X10 of the central metal element, the cohesive energy X11 of the central metal atom, the first dissociation energy X12 of the central metal, the Pauling electronegativity X13 of the central metal, the work function X14 of the central metal, and the melting point X15 of the central metal.

[0016] Preferably, in step S3, the ratio of the test set to the training set is (6 - 9):(1 - 4).

[0017] Preferably, in step S3, the machine learning model includes at least one of a neural network model, a support vector machine model, a random forest model, a decision tree model, and a naive Bayes model.

[0018] Preferably, in step S3, the functional relationship is:

[0019] Y = F(X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14, X15)

[0020] where Y is the limiting oxygen index of the MOFs-modified polyurethane material, and X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14, X15 are the characteristic attribute values of the MOFs-modified polyurethane material.

[0021] Preferably, step S3 further includes: after obtaining the functional relationship between the limiting oxygen index and the characteristic attribute values, calculating the accuracy of the test set and the accuracy of the training set;

[0022] When the accuracy of the test set and the accuracy of the training set are both > 0.8, then step S4 is performed;

[0023] When at least one of the accuracy of the test set and the accuracy of the training set is ≤ 0.8, the parameters are adjusted and retrained to obtain a new functional relationship.

[0024] Preferably, the accuracy of the test set and the accuracy of the training set are measured by the correlation coefficient R 2 metric;

[0025] where

[0026] i represents the number of MOFs-modified polyurethane materials, Y i represents the limiting oxygen index of the MOFs-modified polyurethane material determined by experiment, y i represents the predicted limiting oxygen index of the MOFs-modified polyurethane material obtained according to the functional relationship, represents the average value of the limiting oxygen indices of i MOFs-modified polyurethane materials determined by experiment.

[0027] Preferably, step S4 further includes: verifying the accuracy of the prediction result by experimentally determining the limiting oxygen index of the MOFs-modified polyurethane material with the highest limiting oxygen index screened out.

[0028] The principle of the present invention is as follows: A small portion of MOFs-modified polyurethane materials are randomly prepared through chemical synthesis, their limiting oxygen indices are measured, and the characteristic attribute values of the MOFs-modified polyurethane materials are selected to form a database for machine learning model training. Then, various machine learning models are trained and tested based on the established database, and the limiting oxygen indices of most MOFs-modified polyurethane materials are predicted through various models, so as to screen out the MOFs-modified polyurethane materials with the highest limiting oxygen index. Based on the prediction results, several materials with relatively high limiting oxygen indices are selected for synthesis, and then their limiting oxygen indices are measured through experiments, thereby verifying the accuracy of machine learning and achieving the purpose of developing modified polyurethane foaming materials with excellent performance.

[0029] Compared with the prior art, the present invention has at least the following beneficial effects:

[0030] (1) The present invention is based on machine learning to assist in the development of MOFs-modified polyurethane materials, evaluating the flame retardant performance of materials from tens of thousands of materials. The database is huge, and the selected materials are persuasive.

[0031] (2) The present invention combines machine learning for material development, saving a large amount of economic and time costs brought by experimental trial and error.

[0032] (3) The flame retardant effect of the MOFs-modified polyurethane materials screened by the method described in the present invention is excellent, and the limiting oxygen index is relatively high. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a schematic flow chart of the development of MOFs-modified polyurethane materials based on machine learning assistance described in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] The following will describe in detail the specific embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.

[0035] In the ranges disclosed herein, the endpoints and any value are not limited to the exact range or value. These ranges or values should be understood to include values close to these ranges or values. For numerical ranges, the values between the endpoints of each range, between the endpoints of each range and individual point values, and between individual point values can be combined with each other to obtain one or more new numerical ranges, and these numerical ranges should be regarded as specifically disclosed herein.

[0036] The method for developing MOFs-modified polyurethane materials based on machine learning assistance provided by the present invention includes the following steps:

[0037] S1. Randomly select a variety of MOF-modified polyurethane materials, and measure their limiting oxygen index through experiments; determine the characteristic attribute values of the selected MOF-modified polyurethane materials;

[0038] S2. Construct a database with the limiting oxygen index and the determined characteristic attribute values;

[0039] S3. Divide the database into a test set and a training set, and use a variety of machine learning models for training to obtain the functional relationship between the limiting oxygen index and the characteristic attribute values;

[0040] S4. Predict the limiting oxygen index of various MOF-modified polyurethane materials according to the functional relationship, and screen out the MOF-modified polyurethane material with the highest limiting oxygen index.

[0041] In the method of the present invention, the purpose of step S1 is to randomly select a variety of MOF-modified polyurethane materials for synthesis, then measure the limiting oxygen index of these randomly selected MOF-modified polyurethane materials through experiments, and then determine the characteristic attribute values of these MOF-modified polyurethane materials to provide data for constructing a database.

[0042] In step S1, in order to improve the accuracy of the training results of the machine learning model and the accuracy of the prediction results, the number of randomly selected MOF-modified polyurethane materials cannot be too small. In some embodiments, the number of MOF-modified polyurethane materials randomly selected in step S1 can be 80-120 kinds, preferably 100 kinds.

[0043] In the present invention, the preparation processes of various MOF-modified polyurethane materials are basically the same. In one embodiment, the preparation process of the MOF-modified polyurethane material includes: under the presence of a mixed gas containing argon and oxygen, perform plasma treatment on the MOF material, then disperse the plasma-treated MOF material in an organic solvent for ultrasonic treatment to obtain solution A, then dissolve the polyurethane in the organic solvent to obtain solution B, then mix solution A and solution B, and finally freeze-dry the mixture to obtain the MOF-modified polyurethane material.

[0044] In a more specific embodiment, the preparation process of the MOF-modified polyurethane material includes the following steps:

[0045] (1) Place the MOF material in a plasma treatment chamber, introduce a mixed gas of argon and oxygen, and perform plasma treatment at a certain power and treatment time;

[0046] (2) Disperse the plasma-treated MOF material in an organic solvent for ultrasonic treatment to make it uniformly dispersed to obtain solution A;

[0047] (3) Dissolve the polyurethane in an organic solvent to obtain Liquid B;

[0048] (4) Mix Liquid A and Liquid B and stir in a magnetic stirrer to obtain a mixed solution;

[0049] (5) Freeze-dry the mixed solution to obtain the MOFs-modified polyurethane material.

[0050] In the present invention, the MOF material can be a conventional choice in the art and can be represented as MxLy, where M is the metal center, L is the organic ligand, and x and y are the ratios of the central metal and the organic ligand, respectively. Specifically, M can be a central metal such as Cu, Zn, Fe, Co, Ni, Mn, Cr, V, Ti, Sc, etc., and L can be an organic ligand such as alkyl, carbonyl, alkynyl, carboxyl, amino, hydroxyl, etc. M and L can be randomly combined. In the present invention, the polyurethane can be a conventional choice in the art.

[0051] In the present invention, during the preparation process of various MOFs-modified polyurethane materials, the power and treatment time of the plasma treatment in step (1) need to be determined according to the specific MOFs-modified polyurethane materials. The other conditions of the preparation process of various MOFs-modified polyurethane materials are basically the same.

[0052] In some embodiments, the flow rate of the mixed gas in step (1) is 30-80 mL / min, and the flow rate ratio of argon to oxygen in the mixed gas is 1:1.5-3.

[0053] In some embodiments, the time of the ultrasonic treatment in step (2) can be 0.5-2 hours.

[0054] In some embodiments, the dosage ratio of the MOF material to the polyurethane can be 1:(2-20).

[0055] In some embodiments, the stirring time in step (4) can be 1-5 h, preferably 2 h.

[0056] In some embodiments, the temperature of the freeze-drying in step (5) is -50°C to -20°C, and the time is 48-72 h, preferably 56 h.

[0057] In the present invention, the method for experimentally determining the limiting oxygen index of MOF-modified polyurethane materials includes: processing polyurethane foam materials into standard test specimens (with a width b of about 10 mm, a height h of about 3 mm, and a length l of about 125 mm), and measuring the limiting oxygen index LOI value of the specimens using an oxygen index measuring instrument. Fix the specimen to be tested on the combustion rack, adjust the flow rates of N2 and O2 to a certain fixed value, ignite the specimen to be tested, and observe whether the specimen to be tested can self-extinguish within 3 minutes. If it cannot self-extinguish, increase the flow rate of N2; if it cannot be ignited, increase the flow rate of O2. Repeat the test in this way until self-extinguishing to determine the limiting oxygen index of the specimen. The higher the limiting oxygen index, the better the flame retardancy performance.

[0058] In the present invention, in order to improve the universality and accuracy of the prediction results, the correlation between the characteristic attribute values of the selected MOF-modified polyurethane materials should not be too large. Therefore, when determining the characteristic attribute values in the database, it is necessary to calculate the correlation between multiple selected characteristic attribute values and exclude inappropriate characteristic attribute values.

[0059] In one embodiment, in step S1, the process of determining the characteristic attribute values of the selected MOF-modified polyurethane materials includes: selecting N characteristic attribute values of the MOF-modified polyurethane materials, and calculating the pearson correlation coefficient between any two characteristic attribute values, for a total of calculations; when the pearson correlation coefficient between two characteristic attribute values > 0.9, it indicates that the correlation between the two characteristic attribute values is relatively large, then select one of the two characteristic attribute values to be included in the database and remove the other; when the pearson correlation coefficient between two characteristic attribute values ≤ 0.9, it indicates that the correlation between the two characteristic attribute values is not large, then both characteristic attribute values are included in the database.

[0060] In the present invention, finally, 15 characteristic attribute values of the MOF-modified polyurethane materials are determined as the data for constructing the database. In one embodiment, in step S1, the determined characteristic attribute values of the selected MOF-modified polyurethane materials include the molecular weight X1 of the organic ligand, the number of oxygen atoms X2 of the organic ligand, the number of hydrogen atoms X3 of the organic ligand, the number of hydrogen atoms X4 of the organic ligand, the number of nitrogen atoms X5 of the organic ligand, the relative atomic mass X6 of the central metal, the atomic number X6 of the central metal, the number of electrons in the d orbital X7 of the central metal, the coordination number X8 of the central metal, the main group number X9 of the central metal element, the main group number X10 of the central metal element, the cohesive energy X11 of the central metal atom, the first dissociation energy X12 of the central metal, the Pauling electronegativity X13 of the central metal, the work function X14 of the central metal, and the melting point X15 of the central metal.

[0061] In the present invention, a variety of randomly selected MOF-modified polyurethane materials are divided into two groups according to a certain ratio. The limiting oxygen index and characteristic attribute values corresponding to the first group constitute the test set, and the limiting oxygen index and characteristic attribute values corresponding to the second group constitute the training set. In one embodiment, in step S3, the ratio of the test set to the training set is (6 - 9):(1 - 4).

[0062] In the present invention, the process of training using a variety of machine learning models can be a conventional operation in the art. In some embodiments, the machine learning model in step S3 includes at least one of a neural network model, a support vector machine model, a random forest model, a decision tree model, and a naive Bayes model. After training with the machine learning model, a functional relationship between the limiting oxygen index and the characteristic attribute values can be obtained, where the functional relationship is:

[0063] Y = F(X1,X2,X3,X4,X5,X6,X7,X8,X9,X10,X11,X12,X13,X14,X15)

[0064] where Y is the limiting oxygen index of the MOF-modified polyurethane material, and X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14, X15 are the 15 characteristic attribute values of the MOF-modified polyurethane material described above.

[0065] In the present invention, after training with different machine learning models, the specific functional relationships obtained are slightly different, and ultimately the MOF-modified polyurethane materials with the highest predicted limiting oxygen index selected are different. In the present invention, after training with various machine learning models, the specific functional relationships obtained are all exponential functions, and the exponential function is where a is a coefficient with a range of -5 to 5; m is an exponent with a range of -1 to 1; Y is the limiting oxygen index of the i-th MOF-modified polyurethane material; Xi is the characteristic attribute value of the i-th MOF-modified polyurethane material; N is the number of characteristic attribute values of the MOF-modified polyurethane material. The difference in the exponential functions obtained after training with different machine learning models lies in the different values of a and m.

[0066] To improve the effect of machine learning and ensure the accuracy of prediction results, after obtaining the functional relationship between the limiting oxygen index and the characteristic attribute values, it is necessary to verify this functional relationship. Specifically, it can be verified by calculating the accuracy of the test set and the training set. In one implementation, after obtaining the functional relationship between the limiting oxygen index and the characteristic attribute values, calculate the accuracy of the test set and the training set; when the accuracy of both the test set and the training set > 0.8, it indicates that the functional relationship is reliable, and proceed to step S4; when at least one of the accuracy of the test set and the training set ≤ 0.8, it indicates that the functional relationship is unreliable, then adjust the parameters and retrain to obtain a new functional relationship.

[0067] In a specific implementation, taking the support vector machine model as an example, the process of adjusting the parameters and retraining mainly adjusts two parameters, namely the penalty parameter C and the kernel function parameter m. C controls the degree of penalty for misclassified samples, and its value range is between 2 -5 and 2 15 . The kernel function parameter m affects the distribution of data in the feature space, and its general value range is between 2 -15 and 2 3 . By traversing the parameter space, the optimal parameter combination is determined according to the evaluation index. When a set of parameters is found such that the performance of the model on the validation set reaches the optimum, this set of parameters is the optimal parameter.

[0068] In one implementation, the accuracy of the test set and the training set is measured by the correlation coefficient R 2 ;

[0069] where

[0070] i represents the number of MOFs-modified polyurethane materials, Y i represents the limiting oxygen index measured experimentally for the MOFs-modified polyurethane materials, y i represents the predicted limiting oxygen index of the MOFs-modified polyurethane materials obtained according to the functional relationship, represents the average value of the limiting oxygen indices measured experimentally for i MOFs-modified polyurethane materials.

[0071] In the present invention, when an accurate functional relationship is obtained, the limiting oxygen index of most MOFs-modified polyurethane materials can be predicted based on the characteristic attribute values of tens of thousands of MOFs-modified polyurethane materials, so as to screen out the MOFs-modified polyurethane materials with the highest limiting oxygen index.

[0072] In one embodiment, after screening out the MOFs-modified polyurethane materials with the highest limiting oxygen index, the limiting oxygen index of the selected MOFs-modified polyurethane materials with the highest limiting oxygen index can be experimentally determined to verify the accuracy of the prediction results. Specifically, by calculating the error between the predicted limiting oxygen index and the experimentally determined limiting oxygen index, when the error ≤ 5%, it indicates that the result of machine learning is accurate; when the error > 5%, it indicates that the result of machine learning is inaccurate, and then return to step S2 to reconstruct the database.

[0073] The method described in the present invention first prepares a small part of MOFs-modified polyurethane materials by chemical synthesis, measures their limiting oxygen index, and constitutes a database for machine learning model training. Then, based on the established database, various machine learning models are trained and tested. According to the training results, the limiting oxygen index of most MOFs-modified polyurethane materials is predicted. Based on the predicted results, materials with higher limiting oxygen index are selected for synthesis, and then their limiting oxygen index is experimentally determined to verify the accuracy of machine learning and achieve the purpose of developing modified polyurethane foaming materials with excellent performance. The present invention combines machine learning for material development, saving a large amount of economic and time costs brought by experimental trial and error.

[0074] The present invention will be described in detail below through examples, and the protection scope of the present invention is not limited thereto. In the following examples and comparative examples, unless otherwise specified, all raw materials are common raw materials on the market, and the methods used are conventional methods.

[0075] In the following examples and comparative examples, the polyurethane has a weight average molecular weight of about 100,000 and is purchased from Hubei Qifei Pharmaceutical Chemical Co., Ltd.

[0076] Example 1

[0077] 1) Place Cu1(COOH)2 in a plasma treatment chamber, and introduce a mixed gas of 0.5 mol / L argon and 1 mol / L oxygen at a speed of 50 mL / min for plasma treatment at a power of 100 W for 1 hour;

[0078] 2) Disperse 0.5 g of plasma-treated Cu1(COOH)2 in 30 mL of tetrahydrofuran solvent, and ultrasonically treat it for 1 hour to make it uniformly dispersed to obtain solution A;

[0079] 3) Dissolve 2 g of polyurethane in 20 mL of tetrahydrofuran solvent to obtain solution B;

[0080] 4) Mix solution A and solution B, and stir in a magnetic stirrer for 2 h to obtain a mixed solution;

[0081] 5) Freeze-dry the mixture for 56 h at a temperature of -30 °C to obtain the Cu1(COOH)2 modified polyurethane material, and measure its limiting oxygen index to be 28.7%;

[0082] 6) Replace the MOFs material so that M in MxLy is a central metal such as Cu, Zn, Fe, Co, Ni, Mn, Cr, V, Ti, Sc, etc., and L is an organic ligand such as alkyl, carbonyl, alkynyl, carboxyl, amino, hydroxyl, etc. M and L are randomly combined, and prepare 99 other MOFs modified polyurethane materials according to steps 1)-5), and measure their limiting oxygen indices;

[0083] 7) Determine that the 15 characteristic attribute values of the MOFs modified polyurethane material are respectively the molecular weight X1 of the organic ligand, the number of oxygen atoms X2 of the organic ligand, the number of hydrogen atoms X3 of the organic ligand, the number of hydrogen atoms X4 of the organic ligand, the number of nitrogen atoms X5 of the organic ligand, the relative atomic mass X6 of the central metal, the atomic number X6 of the central metal, the number of electrons in the d orbital X7 of the central metal, the coordination number X8 of the central metal, the main group number X9 of the central metal element, the main group number X10 of the central metal element, the cohesive energy X11 of the central metal atom, the first dissociation energy X12 of the central metal, the Pauling electronegativity X13 of the central metal, the work function X14 of the central metal, and the melting point X15 of the central metal;

[0084] 8) Construct a database with the limiting oxygen indices and 15 characteristic attribute values of 100 MOFs modified polyurethane materials, and divide the database into a test set and a training set. Among them, the ratio of the test set to the training set is 7:3. Then, train using a neural network model based on the established database to obtain an exponential function relationship between the limiting oxygen index and the characteristic attribute values;

[0085] 9) Predict the limiting oxygen indices of various MOFs modified polyurethane materials according to the functional relationship, and screen out the MOFs modified polyurethane material with the highest limiting oxygen index as the Cu1(COOH)2 modified polyurethane material. The predicted limiting oxygen index of the Cu1(COOH)2 modified polyurethane material is 30.1%. Calculate the error between the predicted limiting oxygen index and the measured limiting oxygen index to be 4.98%, which is less than 5%.

[0086] Example 2

[0087] It was implemented according to the method of Example 1, except that in step 8), a support vector machine model was used for training, and the MOFs-modified polyurethane material with the highest limiting oxygen index was selected as the Zn1(COOH)1-modified polyurethane material (the measured limiting oxygen index was 46.7), and the predicted limiting oxygen index of the Zn1(COOH)1-modified polyurethane material was 48.7%. By calculating, the error between the predicted limiting oxygen index and the measured limiting oxygen index was 4.3%, which was less than 5%.

[0088] Example 3

[0089] It was implemented according to the method of Example 1, except that in step 8), a random forest model was used for training, and the MOFs-modified polyurethane material with the highest limiting oxygen index was selected as the Zn1(COOH)2-modified polyurethane material (the measured limiting oxygen index was 46.7%), and the predicted limiting oxygen index of the Zn1(COOH)2-modified polyurethane material was 44.9%. By calculating, the error between the predicted limiting oxygen index and the measured limiting oxygen index was 3.9%, which was less than 5%.

[0090] Example 4

[0091] It was implemented according to the method of Example 1, except that in step 8), a random forest model was used for training, and the MOFs-modified polyurethane material with the highest limiting oxygen index was selected as the Fe1(COOH)2-modified polyurethane material (the measured limiting oxygen index was 36.2%), and the predicted limiting oxygen index of the Fe1(COOH)2-modified polyurethane material was 37.8%. By calculating, the error between the predicted limiting oxygen index and the measured limiting oxygen index was 4.4%, which was less than 5%.

[0092] Example 5

[0093] It was implemented according to the method of Example 1, except that in step 8), a naive Bayes model was used for training, and the MOFs-modified polyurethane material with the highest limiting oxygen index was selected as the Fe1(COOH)2-modified polyurethane material (the measured limiting oxygen index was 36.2%), and the predicted limiting oxygen index of the Fe1(COOH)2-modified polyurethane material was 34.9%. By calculating, the error between the predicted limiting oxygen index and the measured limiting oxygen index was 3.59%, which was less than 5%.

[0094] Comparative Example 1

[0095] Prepare the Cu1(COOH)2-modified polyurethane material and experimentally determine the limiting oxygen index of the Cu1(COOH)2-modified polyurethane material.

[0096] The preparation process of the Cu1(COOH)2-modified polyurethane material was the same as that of Example 1.

[0097] Comparative Example 2

[0098] The Zn1(COOH)2 modified polyurethane material was prepared and its limiting oxygen index was determined experimentally to be 46.7%.

[0099] The preparation process of the Zn1(COOH)2 modified polyurethane material is as follows:

[0100] 1) Place Zn1(COOH)2 in the plasma treatment chamber, and introduce a mixed gas of 0.5 mol / L argon and 1 mol / L oxygen at a speed of 50 mL / min for plasma treatment at a power of 100 W for 1 hour;

[0101] 2) Disperse 0.5 g of the plasma-treated Zn1(COOH)2 in 30 mL of tetrahydrofuran solvent, and ultrasonically treat it for 1 hour to make it uniformly dispersed to obtain Solution A;

[0102] 3) Dissolve 2 g of polyurethane in 20 mL of tetrahydrofuran solvent to obtain Solution B;

[0103] 4) Mix Solution A and Solution B, and stir in a magnetic stirrer for 2 h to obtain a mixed solution;

[0104] 5) Freeze-dry the mixed solution for 56 h at a freeze-drying temperature of -30 °C to obtain the Zn1(COOH)2 modified polyurethane material.

[0105] Comparative Example 3

[0106] The Fe1(COOH)2 modified polyurethane material was prepared and its limiting oxygen index was determined experimentally to be 36.2%.

[0107] The preparation process of the Fe1(COOH)2 modified polyurethane material is as follows:

[0108] 1) Place Fe1(COOH)2 in the plasma treatment chamber, and introduce a mixed gas of 0.5 mol / L argon and 1 mol / L oxygen at a speed of 50 mL / min for plasma treatment at a power of 100 W for 1 hour;

[0109] 2) Disperse 0.5 g of the plasma-treated Fe1(COOH)2 in 30 mL of tetrahydrofuran solvent, and ultrasonically treat it for 1 hour to make it uniformly dispersed to obtain Solution A;

[0110] 3) Dissolve 2 g of polyurethane in 20 mL of tetrahydrofuran solvent to obtain Solution B;

[0111] 4) Mix Solution A and Solution B, and stir in a magnetic stirrer for 2 h to obtain a mixed solution;

[0112] 5) Freeze-dry the mixture for 56 h at a freeze-drying temperature of -30 °C to obtain the Fe1(COOH)2-modified polyurethane material.

[0113] Comparative Example 4

[0114] The limiting oxygen index of the polyurethane material is determined experimentally to be 10.7%.

[0115] As can be seen from the examples, the use of the machine learning method to predict the limiting oxygen index of the MOF-modified polyurethane material is in good agreement with the experimentally determined results, and the error between the predicted results and the experimentally determined results is small.

[0116] In addition, as can be seen from the results of the examples and comparative examples, the use of MOFs to modify polyurethane results in MOF-modified polyurethane materials with excellent flame retardant properties.

[0117] It should be understood that the parts not detailed in this specification belong to the prior art.

[0118] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited thereto. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solutions of the present invention, including any other suitable combination of each technical feature. These simple modifications and combinations should also be regarded as the content disclosed by the present invention and fall within the protection scope of the present invention.

Claims

1. A method for the development of MOF-modified polyurethane materials assisted by machine learning, characterized in that The method includes the following steps: S1. Randomly select a variety of MOF-modified polyurethane materials, and determine their limiting oxygen indices through experiments; determine the characteristic attribute values of the selected MOF-modified polyurethane materials; S2. Construct a database with the limiting oxygen indices and the determined characteristic attribute values; S3. Divide the database into a test set and a training set, and train using a variety of machine learning models to obtain the functional relationship between the limiting oxygen index and the characteristic attribute values; S4. Predict the limiting oxygen indices of various MOF-modified polyurethane materials according to the functional relationship, and screen out the MOF-modified polyurethane material with the highest limiting oxygen index.

2. The method according to claim 1, wherein In step S1, the number of randomly selected MOF-modified polyurethane materials can be 80 - 120 kinds; And / or, the preparation process of the MOF-modified polyurethane material includes: in the presence of a mixed gas containing argon and oxygen, perform plasma treatment on the MOF material, then disperse the plasma-treated MOF material in an organic solvent for ultrasonic treatment to obtain liquid A, then dissolve the polyurethane in the organic solvent to obtain liquid B, then mix liquid A and liquid B, and finally freeze-dry the mixed solution to obtain the MOF-modified polyurethane material.

3. The method according to claim 1 or 2, characterized in that, In step S1, the process of determining the characteristic attribute values of the selected MOF-modified polyurethane materials includes: Select N characteristic attribute values of the MOF-modified polyurethane material, and calculate the Pearson correlation coefficient between any two characteristic attribute values, with a total of times; When the pearson correlation coefficient of two characteristic attribute values > 0.9, then select one of the two characteristic attribute values to enter the database; When the pearson correlation coefficient of two characteristic attribute values ≤ 0.9, then both characteristic attribute values enter the database.

4. The method according to claim 1, wherein In step S1, the characteristic attribute values determined for the selected MOF-modified polyurethane materials include the molecular weight X1 of the organic ligand, the number of oxygen atoms X2 of the organic ligand, the number of hydrogen atoms X3 of the organic ligand, the number of hydrogen atoms X4 of the organic ligand, the number of nitrogen atoms X5 of the organic ligand, the relative atomic mass X6 of the central metal, the atomic number X6 of the central metal, the number of electrons in the d orbital X7 of the central metal, the coordination number X8 of the central metal, the main group number X9 of the central metal element, the main group number X10 of the central metal element, the cohesive energy X11 of the central metal atom, the first dissociation energy X12 of the central metal, the Pauling electronegativity X13 of the central metal, the work function X14 of the central metal, and the melting point X15 of the central metal.

5. The method according to claim 1, wherein In step S3, the ratio of the test set to the training set is (6 - 9):(1 - 4).

6. The method according to claim 1, wherein In step S3, the machine learning model includes at least one of a neural network model, a support vector machine model, a random forest model, a decision tree model, and a naive Bayes model.

7. The method according to claim 1 or 4, characterized in that, In step S3, the functional relationship is: Y = F(X1,X2,X3,X4,X5,X6,X7,X8,X9,X10,X11,X12,X13,X14,X15) Among them, Y is the limiting oxygen index of the MOFs modified polyurethane material, and X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14, X15 are the characteristic attribute values of the MOFs modified polyurethane material.

8. The method according to claim 1, characterized in that Step S3 also includes: after obtaining the functional relationship between the limiting oxygen index and the characteristic attribute values, calculating the accuracy of the test set and the accuracy of the training set; When the accuracy of the test set and the accuracy of the training set are both > 0.8, then step S4 is performed; When at least one of the accuracy of the test set and the accuracy of the training set is ≤ 0.8, then the parameters are adjusted and retrained to obtain a new functional relationship.

9. The method according to claim 8, wherein The accuracy of the test set and the accuracy of the training set are measured by the correlation coefficient R 2 metric; Among them, i represents the number of MOF-modified polyurethane materials, Y i represents the limiting oxygen index measured experimentally for the MOF-modified polyurethane materials, y i represents the predicted limiting oxygen index of the MOF-modified polyurethane materials obtained from the functional relationship represents the average value of the limiting oxygen indices measured experimentally for i MOF-modified polyurethane materials.

10. The method according to claim 1, characterized in that, Step S4 also includes: experimentally determining the limiting oxygen index of the MOFs modified polyurethane material with the highest limiting oxygen index selected, and verifying the accuracy of the prediction result.

Images