Method for developing MOFs modified polypropylene material based on high-throughput screening and machine learning
By combining high-throughput screening with machine learning, MOFs-modified polypropylene materials with excellent high and low temperature resistance were screened out, solving the problem of polypropylene materials being easily broken at extreme temperatures and achieving stable use and cost savings of MPP pipelines.
Patent Information
- Application Number
- CN202510567386.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-30
- Publication Date
- 2025-09-16
AI Technical Summary
Polypropylene materials have insufficient performance at low and high temperatures, and the development efficiency of existing MOFs modified materials is low, resulting in pipes that are prone to rupture at extreme temperatures and are costly.
A method combining high-throughput screening and machine learning was used to synthesize MOFs-modified polypropylene materials, measure characteristic values and perform standardization, construct an initial database, and use machine learning models to predict performance parameters to screen out MOFs-modified polypropylene materials that meet the requirements.
The high and low temperature resistance of polypropylene materials has been significantly improved, the experimental costs have been reduced, and the stable use of MPP pipes in a wide temperature range has been achieved.
Smart Images
Figure CN120656595A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of functional materials, and in particular to a method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning. Background Art
[0002] MPP pipes are important in the power, communications, municipal services, agricultural irrigation, and other fields. These pipes are required to exhibit high and low temperature resistance, a high heat deformation temperature and low-temperature impact resistance, and the ability to operate properly over a wide temperature range. Polypropylene is primarily used as the raw material for these pipes. However, polypropylene's low-temperature resistance is relatively weak, with its glass transition temperature ranging from approximately -10°C to 0°C. Below this temperature, polypropylene's flexibility and impact toughness decrease significantly, becoming hard and brittle, making it susceptible to fracture when subjected to external forces. Furthermore, polypropylene's high-temperature resistance is poor, with a melting point typically around 160°C-170°C. At temperatures above 200°C, polypropylene's mechanical properties and dimensional stability significantly decline. Therefore, modifying polypropylene to improve its high- and low-temperature resistance is essential.
[0003] Metal-organic frameworks (MOFs) are porous materials formed by the self-assembly of inorganic metal centers and organic ligands through bridging. They possess large surface areas, excellent thermal and structural stability, and these characteristics meet the fundamental requirements for their use as modifiers. However, MOFs have a wide variety of metal centers and organic ligands. Different metal ions and organic ligands result in MOFs with varying structures, pore sizes, and functions, resulting in tens of thousands of different MOF types.
[0004] However, it is currently unclear which MOFs should be used to modify polypropylene to impart excellent high- and low-temperature resistance. Repeated trial and error experiments consume significant manpower and material resources. Therefore, the present invention provides a method for developing MOF-modified polypropylene materials based on high-throughput screening and machine learning, which is of great significance for rapidly screening MOF-modified polypropylene with excellent high- and low-temperature resistance. Summary of the Invention
[0005] To address the existing challenges of polypropylene materials' insufficient high- and low-temperature resistance and the inefficient development of MOF-modified materials, this paper provides a method for developing MOF-modified polypropylene materials based on high-throughput screening and machine learning. By integrating materials science with artificial intelligence, this method enables the rapid prediction and precise synthesis of high-performance modified materials, significantly reducing experimental trial-and-error costs.
[0006] To achieve the above purpose, the specific technical solutions of the present invention are as follows:
[0007] In a first aspect, the present invention provides a method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning, comprising the following steps:
[0008] S1. Synthesizing a series of MOFs-modified polypropylene materials, respectively measuring characteristic values and performance parameters of the MOFs-modified polypropylene materials, and standardizing the characteristic values to obtain standardized characteristic values;
[0009] S2. Based on the Pearson correlation coefficient of the standardized eigenvalues, an initial database including standardized eigenvalues and performance parameters is constructed;
[0010] S3. Using the data from the initial database to train and test different pairs of machine learning models, and screen out the optimal machine learning model; wherein, using the data from the initial database to train different pairs of machine learning models, a function of performance parameters and standardized eigenvalues is constructed;
[0011] S4. Use the optimal machine learning model to predict the performance parameters of potential MOFs-modified polypropylene materials and screen out MOFs-modified polypropylene materials that meet the requirements.
[0012] Furthermore, the method also includes: using the optimal machine learning model to predict the performance parameters of potential MOFs-modified polypropylene materials, selecting the top 5%-10% of the predicted MOFs-modified polypropylene materials for synthesis, and measuring their performance parameters; comparing the relative errors between the measured values and the predicted values, and the MOFs-modified polypropylene materials with a relative error of ≤5% are considered to meet the requirements; otherwise, repeating steps S3-S4 until the relative error is ≤5%.
[0013] Furthermore, in step S1, the characteristic values include but are not limited to the molecular weight of the organic ligand, the number of oxygen atoms in the organic ligand, the number of hydrogen atoms in the organic ligand, the number of carbon atoms in the organic ligand, the number of nitrogen atoms in the organic ligand, the relative atomic mass of the central metal, the atomic number (AN) of the central metal, the number of electrons in the d orbital of the central metal, the coordination number of the central metal, the main group number of the central metal, the atomic cohesive energy of the central metal, the first dissociation energy of the central metal, the Pauling electronegativity of the central metal, the work function of the central metal and the melting point of the central metal; the performance parameters include but are not limited to the heat deformation temperature and the low-temperature brittle temperature.
[0014] Furthermore, in step S1, the formula for normalizing the characteristic value is as follows:
[0015]
[0016] Wherein, X is the standardized characteristic value; x is the original characteristic value of MOFs modified polypropylene material; and are the mean and mean square error of the original eigenvalues, respectively.
[0017] Furthermore, in step S2, the Pearson correlation coefficient ( P ) is calculated as follows:
[0018]
[0019] in, M i is the standardized characteristic value M of the i-th material in a series of MOFs modified polypropylene materials, is the average value of the standardized characteristic value M in a series of MOFs modified polypropylene materials, that is, =( M 1 +M 2 +……+M n ) / n; N i is the standardized characteristic value N of the i-th material in a series of MOFs modified polypropylene materials, is the average value of the standardized characteristic value N in a series of MOFs modified polypropylene materials, that is, =( N 1 +N 2 +……+N n ) / n; the value of P ranges from -1.0 to 1.0. The higher the absolute value of P, the stronger the correlation.
[0020] Furthermore, in step S2, the standardized eigenvalues and performance parameters with a Pearson correlation coefficient less than 0.9 are used to construct an initial database. When the Pearson correlation coefficient is greater than 0.9, it indicates that the correlation between the two standardized eigenvalues is too large, and the independence between the corresponding two original eigenvalues is not strong. The present invention selects standardized eigenvalues and performance parameters with a Pearson correlation coefficient less than 0.9 to construct an initial database so that in subsequent steps, functions can be constructed using standardized eigenvalues and performance parameters with strong independence.
[0021] Furthermore, in step S3, the different machine learning models include but are not limited to neural networks, random forests, support vector machines (SVMs), decision trees, and naive Bayes models.
[0022] Furthermore, in step S3, the data in the initial database is divided into a training set and a test set. The training set is used to train different machine learning models, and the test set is used to test different machine learning models. The machine learning model with an accuracy greater than 0.8 in both the test set and the training set is screened out, which is the optimal machine learning model.
[0023] Furthermore, the accuracy of the test set and the training set is calculated using the correlation coefficient ( R 2 ) is measured and the calculation formula is as follows:
[0024]
[0025] in, Y i is the actual measured performance parameter; y i is the performance parameter predicted by the machine learning model; It is the average value of the actual measured performance parameters.
[0026] Furthermore, in step S3, the function of constructing performance parameters and standardized feature values through the machine learning model can be expressed as:
[0027] Assume the performance parameter is Y , where the heat deformation temperature is Y1 , low temperature brittle temperature is Y2 ; The standardized characteristic value is X (include X1, X2, X3...X15 ),but:
[0028] Y1 = F( X1, X2, X3...X1 5)
[0029] Y2 = F( X1, X2, X3...X15 )
[0030] Furthermore, in step S3, the function of constructing performance parameters and standardized feature values through the machine learning model is as follows:
[0031]
[0032] in, Y is the performance parameter; a, b, c is the coefficient, a The value ranges from -5 to 5. b The value ranges from -1 to 1. c The value of is -1~1; X i For the i Standardized characteristic values of MOFs modified polypropylene materials; nis the number of standardized eigenvalues. The difference between the functions obtained after training with different machine learning models is a, b, c different.
[0033] Furthermore, in step S4, the heat deformation temperature of the MOFs-modified polypropylene material that meets the requirements is higher than that of the unmodified polypropylene by more than 120°C, and the low-temperature brittle temperature is lower than that of the unmodified polypropylene by more than 25°C. The MOFs-modified polypropylene material that meets the requirements is used to prepare an MPP pipe, and the MPP pipe has an impact strength of ≥30 kJ / m at a temperature of -50-200°C. 2 , thermal deformation ≤2%.
[0034] Furthermore, in the MOFs-modified polypropylene material, the structure of MOFs is MxLy, wherein M is a metal center, L is an organic ligand, and x and y are the molar ratios of M and L, respectively.
[0035] Furthermore, the metal center includes but is not limited to Zn, Cu, Ni, Co, Fe, Cr, Mn, V, Ti, Sc or Fe; the organic ligand includes but is not limited to one or more of -COOH, -OH, -NH2, -CH3, -CH2COOH, -CH3CH2COOH, -C2H, -C3H3, and -C4H5.
[0036] Furthermore, the synthesis method of the MOFs-modified polypropylene material comprises the following steps:
[0037] P1. MOFs were placed in a plasma treatment chamber and treated with a mixture of argon and oxygen at a specific power and treatment time.
[0038] P2. Dispersing the plasma-treated MOFs in an organic solvent and ultrasonically dispersing them uniformly to obtain Solution A.
[0039] P3. Dissolving polypropylene in an organic solvent to obtain solution B;
[0040] P4. Mix Liquid A and Liquid B evenly and freeze-dry to obtain MOFs-modified polypropylene material.
[0041] Furthermore, the mass ratio of the polypropylene to MOFs is (2-10):1.
[0042] Furthermore, in step P1, the plasma treatment conditions include: an argon to oxygen ratio of 1:(3-6), a power of 50-200 W, a treatment time of 0.02-0.1 h, and a particle size of the treated MOFs particles ≤100 nm.
[0043] Furthermore, the plasma treatment conditions include: a ratio of argon to oxygen of 1:5, a power of 100 W, and a treatment time of 0.05 h.
[0044] Furthermore, the organic solvent in step P2 includes but is not limited to one or more of toluene, tetrahydrofuran, and ethyl acetate.
[0045] Furthermore, the mixing in step P4 is carried out in a magnetic stirrer for 1-5 h.
[0046] Furthermore, the mixing time in step P4 is 2 h.
[0047] Furthermore, the freeze-drying time in step P4 is 48-72 h.
[0048] Furthermore, the freeze-drying time in step P4 is 56 h.
[0049] Specifically, the method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning includes the following steps:
[0050] Step A1: randomly synthesizing 100 MOFs-modified polypropylene materials, measuring the performance parameters (heat deformation temperature Td and low-temperature brittle temperature Tb) of the MOFs-modified polypropylene materials using a differential scanning calorimeter and a low-temperature impact tester, and statistically calculating 15 characteristic values of the MOFs-modified polypropylene materials, and then standardizing the characteristic values;
[0051] Step A2: Calculate the Pearson correlation coefficient of each standardized eigenvalue ( P );
[0052] Decision point D1: If P <0.9, then enter the machine learning model training branch, otherwise return to calculate the Pearson correlation coefficient of the standardized eigenvalue;
[0053] Step A3: P An initial database was constructed using standardized eigenvalues and performance parameters with a value < 0.9. The data in the initial database was divided into a 70% training set and a 30% test set. The training set was used to iteratively train neural networks, random forests, support vector machines, decision trees, and naive Bayes models in parallel.
[0054] Decision point D2: Verify the double precision of each machine learning model, the accuracy R of the training set and the test set 2If both are greater than 0.8, the optimal machine learning model is screened. For machine learning models that do not meet the criteria, return to step A3 to adjust parameters. For support vector machine models, adjust the penalty parameter C. A larger C indicates a tendency for the model to overfit, while a smaller C indicates a tendency for the model to underfit. Therefore, adjust C based on actual conditions. For random forest models, adjust the number and depth of trees. Increasing the number and depth of trees improves fitting but may lead to overfitting, so choose a reasonable number and depth of trees.
[0055] Step A4: Using the best machine learning model to predict the performance parameters of thousands of potential MOF-modified polypropylene materials (i.e., MOF-modified polypropylene materials with known characteristic values but unknown performance parameters), screen candidate MOF-modified polypropylene materials with Td ≥ 250 °C and Tb ≤ -30 °C;
[0056] Step A5: Synthesize the MOFs-modified polypropylene materials ranked in the top 5%-10% and measure their performance parameters;
[0057] Decision point D3: Compare the relative errors between the measured and predicted values. If the relative error is ≤5%, the MOF-modified polypropylene material that meets the requirements is output (path Y). Otherwise, the new data is fed back to step A3 and the binary search method is used for iteration to continuously approach the optimal value (path N).
[0058] In a second aspect, the present invention provides a MOFs-modified polypropylene material obtained by the method, wherein the heat deformation temperature of the MOFs-modified polypropylene material is higher by more than 120°C than that of unmodified polypropylene, and the low-temperature brittle temperature is lower by more than 25°C than that of unmodified polypropylene.
[0059] In a third aspect, the present invention provides a computer-readable storage medium storing a program, wherein when the program is executed by a processor, the method is implemented.
[0060] In a fourth aspect, the present invention provides a computer device comprising a processor and a memory for storing a program executable by the processor, wherein the method is implemented when the processor executes the program stored in the memory.
[0061] Compared with the prior art, the present invention is beneficial in that:
[0062] (1) The present invention synthesizes a small number of MOFs-modified polypropylene materials through chemical synthesis, measures their characteristic values and performance parameters, and standardizes the characteristic values. An initial database is constructed based on the Pearson correlation coefficient of the standardized characteristic values and the performance parameters. Various machine learning models are iteratively trained and tested using the initial database, and the best-performing machine learning model is selected to predict the heat deformation temperature and low-temperature brittle temperature of potential MOFs-modified polypropylene materials. Based on the predicted results, several materials with higher heat deformation temperatures and lower low-temperature brittle temperatures are selected for synthesis. The heat deformation temperatures and low-temperature brittle temperatures are then experimentally measured to verify the accuracy of the machine learning, thereby achieving the purpose of developing modified polypropylene materials for MPP pipelines.
[0063] (2) The present invention provides a MOFs-modified polypropylene material, which has excellent high-temperature and low-temperature resistance. Compared with unmodified polypropylene materials, the MOFs-modified polypropylene material has an increased heat deformation temperature and a decreased low-temperature brittle temperature.
[0064] (3) The present invention is based on high-throughput screening to assist in the development of MOFs-modified polypropylene materials. The high-temperature and low-temperature resistance of materials are evaluated from tens of thousands of materials. The database is huge and the screened materials are convincing.
[0065] (4) The present invention combines machine learning with material development, saving a lot of economic and time costs caused by trial and error. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 To develop the heat deformation temperature correlation of MOFs modified polypropylene materials based on high-throughput screening and machine learning;
[0067] Figure 2 The heat deformation temperature of MOFs modified polypropylene material predicted by support vector machine model;
[0068] Figure 3 The low-temperature brittle temperature of MOFs modified polypropylene material predicted by support vector machine model;
[0069] Figure 4 This is an operational flow chart for the present invention to develop MOFs-modified polypropylene materials based on high-throughput screening and machine learning. DETAILED DESCRIPTION
[0070] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0071] The present invention provides a method for developing MOFs modified polypropylene materials based on high throughput screening and machine learning (flow chart as shown in FIG Figure 4 ), including the following steps:
[0072] S1. Synthesizing a series of MOFs-modified polypropylene materials, respectively measuring characteristic values and performance parameters of the MOFs-modified polypropylene materials, and standardizing the characteristic values to obtain standardized characteristic values;
[0073] The characteristic values include but are not limited to the molecular weight of the organic ligand, the number of oxygen atoms in the organic ligand, the number of hydrogen atoms in the organic ligand, the number of carbon atoms in the organic ligand, the number of nitrogen atoms in the organic ligand, the relative atomic mass of the central metal, the atomic number (AN) of the central metal, the number of electrons in the d orbital of the central metal, the coordination number of the central metal, the main group number of the central metal, the atomic cohesive energy of the central metal, the first dissociation energy of the central metal, the Pauling electronegativity of the central metal, the work function of the central metal and the melting point of the central metal; the performance parameters include but are not limited to the heat deformation temperature and the low-temperature brittle temperature;
[0074] The formula for normalizing the eigenvalues is as follows:
[0075]
[0076] in, X is the standardized eigenvalue; x is the original characteristic value of MOFs modified polypropylene material; and are the mean and mean square error of the original eigenvalues, respectively.
[0077] S2. Based on the Pearson correlation coefficient of the standardized eigenvalues, an initial database is constructed by combining the standardized eigenvalues with a Pearson correlation coefficient less than 0.9 and the performance parameters;
[0078] The calculation formula of the Pearson correlation coefficient (P) is as follows:
[0079]
[0080] in, Mi is the standardized characteristic value M of the i-th material in a series of MOFs modified polypropylene materials, is the average value of the standardized characteristic value M in a series of MOFs modified polypropylene materials, that is, =(M1+M2+……+M n ) / n; N i is the standardized characteristic value N of the i-th material in a series of MOFs modified polypropylene materials, is the average value of the standardized characteristic value N in a series of MOFs modified polypropylene materials, that is, =(N1+N2+……+N n ) / n; the value of P ranges from -1.0 to 1.0. The higher the absolute value of P, the stronger the correlation.
[0081] S3. Divide the data in the initial database into a training set and a test set. Use the training set to iteratively train different machine learning models (including but not limited to neural networks, random forests, support vector machines (SVMs), decision trees, and naive Bayesian models) to construct a function of performance parameters and standardized eigenvalues. Use the test set to test the different machine learning models and select the machine learning model with an accuracy greater than 0.8 on both the test set and the training set, which is the optimal machine learning model.
[0082] The function of constructing performance parameters and feature values through machine learning models can be expressed as:
[0083] Assume the performance parameter is Y , where the heat deformation temperature is Y1 , low temperature brittle temperature is Y2 ; Let the eigenvalue be X (include X1, X2, X3...X15 ),but:
[0084] Y1 = F( X1, X2, X3……X15 )
[0085] Y2 = F( X1, X2, X3……X15 )
[0086] The function of constructing performance parameters and standardized feature values through machine learning models is as follows:
[0087]
[0088] in, Y is the performance parameter; a, b, c is the coefficient, a The value ranges from -5 to 5. b The value ranges from -1 to 1. c The value of is -1~1;X i is the standardized characteristic value of the i-th MOFs modified polypropylene material; n is the number of standardized eigenvalues. The difference between the functions obtained after training with different machine learning models is a, b, c different;
[0089] The accuracy of the test set and the training set is calculated using the correlation coefficient ( R 2 ) is measured and the calculation formula is as follows:
[0090]
[0091] in, Y i is the actual measured performance parameter; y i is the performance parameter predicted by the machine learning model; It is the average value of the actual measured performance parameters.
[0092] S4. Use the optimal machine learning model to predict the performance parameters of potential MOFs-modified polypropylene materials, select the top 5%-10% of the predicted MOFs-modified polypropylene materials for synthesis, and measure their performance parameters; compare the relative errors between the measured values and the predicted values, and the MOFs-modified polypropylene materials with a relative error of ≤5% are the ones that meet the requirements; otherwise, repeat steps S3-S4 until the relative error is ≤5%. The heat deformation temperature of the MOFs-modified polypropylene material that meets the requirements is higher than that of the unmodified polypropylene by more than 120°C, and the low-temperature brittle temperature is lower than that of the unmodified polypropylene by more than 25°C. The MOFs-modified polypropylene material that meets the requirements is used to prepare an MPP pipe, and the impact strength of the MPP pipe at a temperature of -50-200°C is ≥30kJ / m 2 , thermal deformation ≤ 2%
[0093] In the following specific examples, the synthesis method of the MOFs modified polypropylene material is as follows:
[0094] P1. Place Cu1(COOH)2 in a plasma treatment chamber and introduce a mixture of 0.5 mol / L argon and 1 mol / L oxygen at a rate of 50 mol / min at a power of 100 W for 1 h.
[0095] P2. Plasma-treated Cu1(COOH)2 was dispersed in tetrahydrofuran and ultrasonicated for 1 h to obtain a uniform dispersion, obtaining solution A.
[0096] P3. Dissolve polypropylene in tetrahydrofuran and stir in an 80°C water bath to obtain Solution B.
[0097] P3. Liquids A and B were mixed and stirred in a magnetic stirrer for 2 hours. The mixture was freeze-dried to obtain a Cu1(COOH)2-modified polypropylene (PP). Its heat deformation temperature and low-temperature brittleness temperature were measured. The mass ratio of MOFs to PP was 1:5.
[0098] P5. Replace MxLy so that M = (Zn, Ni, Co, Fe, Cr, Mn, V, Ti, Sc) and L = (-COOH, -OH, -NH2, -CH3, -CH2COOH, -CH3CH2COOH, -C2H, -C3H3, -C4H5), and repeat S1-S4 multiple times to synthesize a series of MOFs-modified polypropylene materials. Measure 15 characteristic values and performance parameters of the resulting MOFs-modified polypropylene materials. The 15 characteristic values of the MOFs modified polypropylene material are: the molecular weight of the organic ligand, the number of oxygen atoms in the organic ligand, the number of hydrogen atoms in the organic ligand, the number of carbon atoms in the organic ligand, the number of nitrogen atoms in the organic ligand, the relative atomic mass of the central metal, the atomic number of the central metal, the number of electrons in the d orbital of the central metal, the coordination number of the central metal, the main group number of the central metal, the main group number of the central metal, the first dissociation energy of the central metal, the Pauling electronegativity of the central metal, the work function of the central metal and the melting point of the central metal; the performance parameters are the heat deformation temperature and the low-temperature brittle temperature.
[0099] In the following specific examples, the heat deformation temperature test method is as follows: the MOFs-modified polypropylene material to be tested is processed into a standard test specimen (10 mm wide, 3 mm high, and 125 mm long); the specimen is placed in an apparatus, heated at a heating rate of 1°C / min, and a load of 5 kg is applied. The temperature at which the specimen deforms by 2 mm is measured, which is the heat deformation temperature of the material. The low-temperature brittleness temperature test method is as follows: the MOFs-modified polypropylene material to be tested is processed into a standard test specimen (10 mm wide, 3 mm high, and 125 mm long); the specimen is placed in -50°C, -45°C, -40°C, -35°C, -30°C, -25°C, -20°C, -15°C, -10°C, -5°C, and 0°C for 2 hours, then an impact load is applied to the specimen, and the specimen is observed for brittle cracking. The highest temperature at which brittle cracking occurs is determined, which is the low-temperature brittleness temperature of the material.
[0100] In the following examples and comparative examples, unless otherwise specified, all raw materials are commonly available on the market, and all methods used are conventional methods.
[0101] Example 1
[0102] The present invention provides a method for developing MOFs modified polypropylene materials based on high throughput screening and machine learning (see the flow chart) Figure 4 ), including the following steps:
[0103] S1. Randomly synthesize 100 MOFs-modified polypropylene materials, measure the performance parameters (heat deformation temperature Td and low-temperature brittle temperature Tb) of the MOFs-modified polypropylene materials using a differential scanning calorimeter and a low-temperature impact tester, and calculate 15 characteristic values of the MOFs-modified polypropylene materials, which are then standardized.
[0104] S2. Calculate the Pearson correlation coefficient (P) of each standardized eigenvalue.
[0105] S3. Construct an initial database using the standardized eigenvalues and performance parameters with a Pearson correlation coefficient less than 0.9.
[0106] S4. Divide the data in the initial database into training sets and test sets, and use the training sets to iteratively train the neural network, random forest, support vector machine (SVM), decision tree and naive Bayes model respectively, and construct a function of performance parameters and eigenvalues. Let the performance parameter be Y , where the heat deformation temperature is Y1 , low temperature brittle temperature is Y2 ; The standardized characteristic value is X , as shown below:
[0107]
[0108] The function of constructing performance parameters and feature values through machine learning models can be expressed as:
[0109] Y1 = F( X1, X2, X3……X15 )
[0110] Y2 = F( X1, X2, X3……X15 )
[0111] Specifically, the function of constructing performance parameters and standardized feature values through machine learning models is as follows:
[0112]
[0113] in, Y is the performance parameter; a, b, c is the coefficient, a The value ranges from -5 to 5. b The value ranges from -1 to 1. c The value of is -1~1; X i For thei Standardized characteristic values of MOFs modified polypropylene materials; n is the number of standardized eigenvalues. The difference between the functions obtained after training with different machine learning models is a, b, c different.
[0114] Then, the neural network, random forest, support vector machine (SVM), decision tree and naive Bayes models were tested using the test set, that is, the standardized feature values of the potential MOFs modified polypropylene materials were input. X (include X1, X2, X3……X15 ), outputs the performance parameters predicted by the machine learning model Y (Including heat distortion temperature Y1 and low temperature brittle temperature is Y2 ). The machine learning model with an accuracy greater than 0.8 in both the test set and the training set is selected as the optimal machine learning model;
[0115] The accuracy of the test set and the training set is calculated using the correlation coefficient ( R 2 ) is measured and the calculation formula is as follows:
[0116]
[0117] in, Y i is the actual measured performance parameter; y i is the performance parameter predicted by the machine learning model; It is the average value of the actual measured performance parameters.
[0118] Figure 1 Correlation coefficients for different machine learning models in predicting the heat deformation temperature of MOFs-modified polypropylene materials.
[0119] S4. Use the optimal machine learning model to predict the performance parameters of potential MOFs-modified polypropylene materials, select the top 5%-10% of the predicted MOFs-modified polypropylene materials for synthesis, and measure their performance parameters;
[0120] Figure 2 The heat deformation temperature of some MOFs modified polypropylene materials predicted by the optimal machine learning model; Figure 3 The low-temperature brittle temperature of some MOFs-modified polypropylene materials was predicted using the optimal machine learning model.
[0121] Compare the relative errors of the measured values and the predicted values. If the relative error is ≤5%, the MOFs-modified polypropylene material meets the requirements; otherwise, repeat steps S3-S4 until the relative error is ≤5%.
[0122] The MOFs-modified polypropylene material that meets the requirements has a heat deformation temperature that is 120°C higher than that of unmodified polypropylene (i.e., >280°C), and a low-temperature brittle temperature that is 25°C lower than that of unmodified polypropylene (i.e., <-25°C). The MOFs-modified polypropylene material that meets the requirements is used to prepare an MPP pipe, and the MPP pipe has an impact strength of ≥30 kJ / m at a temperature of -50-200°C. 2 , thermal deformation ≤2%.
[0123] The structure and performance parameters of the MOFs-modified polypropylene material that meet the requirements predicted by the method of the present invention are as follows:
[0124]
[0125] As can be seen from the table, the MOFs-modified polypropylene material predicted using the method of the present invention has a heat deformation temperature of over 280°C and a low-temperature brittle temperature below -30°C. This indicates that the method of the present invention uses machine learning to develop MOFs-modified polypropylene materials, and can predict and screen a large number of potential MOFs-modified polypropylene materials, saving a lot of economic and time costs caused by trial and error experiments.
[0126] In summary, the present invention synthesizes a small number of MOFs-modified polypropylene materials through chemical synthesis, determines their characteristic values and performance parameters, and standardizes the characteristic values. An initial database is constructed based on the Pearson correlation coefficient of the standardized characteristic values and the performance parameters. Various machine learning models are trained and tested using the data from the initial database. The machine learning model with the best performance is selected to predict the heat deformation temperature and low-temperature brittle temperature of potential MOFs-modified polypropylene materials, and MOFs-modified polypropylene materials that meet the requirements are screened out, saving a lot of economic and time costs caused by experimental trial and error, and achieving the purpose of quickly developing modified polypropylene materials for MPP pipelines.
[0127] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.
Claims
1. A method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning, characterized in that: The following steps are involved: S1. Synthesizing a series of MOFs-modified polypropylene materials, respectively measuring characteristic values and performance parameters of the MOFs-modified polypropylene materials, and standardizing the characteristic values to obtain standardized characteristic values; S2. constructing an initial database including standardized eigenvalues and performance parameters based on the Pearson correlation coefficient of the standardized eigenvalues; S3. Using the data from the initial database, train and test different pairs of machine learning models to select the optimal machine learning model; wherein, using the data from the initial database, iteratively train the different pairs of machine learning models to construct a function of performance parameters and standardized eigenvalues; S4. Use the optimal machine learning model to predict the performance parameters of potential MOFs-modified polypropylene materials and screen out MOFs-modified polypropylene materials that meet the requirements.
2. The method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning according to claim 1, characterized in that: Also includes: After using the optimal machine learning model to predict the performance parameters of potential MOFs-modified polypropylene materials, the top 5%-10% of the predicted MOFs-modified polypropylene materials were selected for synthesis and their performance parameters were measured. The relative error between the measured and predicted values was compared, and those with a relative error of ≤5% were considered MOFs-modified polypropylene materials that met the requirements. Otherwise, repeat steps S3-S4 until the relative error is ≤ 5%.
3. The method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning according to claim 1, characterized in that: In step S2, an initial database is constructed using the standardized eigenvalues with a Pearson correlation coefficient less than 0.9 and the performance parameters.
4. The method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning according to claim 3, characterized in that: In step S2, the calculation formula of the Pearson correlation coefficient P is as follows: ； in, M i is the standardized characteristic value M of the i-th material in a series of MOFs modified polypropylene materials, is the average value of the standardized characteristic value M in a series of MOFs modified polypropylene materials; N i is the standardized characteristic value N of the i-th material in a series of MOFs modified polypropylene materials, is the average value of the standardized characteristic value N in a series of MOFs modified polypropylene materials.
5. The method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning according to claim 1, characterized in that: In step S3, the data in the initial database is divided into a training set and a test set. The training set is used to iteratively train different machine learning models, and the test set is used to test different machine learning models. The machine learning model with an accuracy greater than 0.8 in both the test set and the training set is selected as the optimal machine learning model.
6. The method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning according to claim 5, characterized in that: The accuracy of the test set and the training set is calculated using the correlation coefficient R 2 The calculation formula is as follows: ； in, Y i is the actual measured performance parameter; y i is the performance parameter predicted by the machine learning model; It is the average value of the actual measured performance parameters.
7. The method for developing MOFs-modified polypropylene materials based on high-throughput screening and machine learning according to claim 1, characterized in that: In step S1, the characteristic values include but are not limited to the molecular weight of the organic ligand, the number of oxygen atoms in the organic ligand, the number of hydrogen atoms in the organic ligand, the number of carbon atoms in the organic ligand, the number of nitrogen atoms in the organic ligand, the relative atomic mass of the central metal, the atomic number of the central metal, the number of electrons in the d orbital of the central metal, the coordination number of the central metal, the main group number of the central metal, the atomic cohesive energy of the central metal, the first dissociation energy of the central metal, the Pauling electronegativity of the central metal, the work function of the central metal and the melting point of the central metal; the performance parameters include but are not limited to the heat deformation temperature and the low-temperature brittle temperature.
8. The MOFs-modified polypropylene material obtained by the method according to any one of claims 1 to 7, characterized in that: The heat deformation temperature of the MOFs-modified polypropylene material is higher than that of unmodified polypropylene by more than 120°C, and the low-temperature brittle temperature is lower than that of unmodified polypropylene by more than 25°C.
9. A computer-readable storage medium storing a program, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
10. A computer device, characterized in that: The invention comprises a processor and a memory for storing a program executable by the processor, wherein when the processor executes the program stored in the memory, the method according to any one of claims 1 to 7 is implemented.