Method, medium and equipment for screening, adsorbing and separating inert gas materials

Abstract

The invention relates to a method, a medium and equipment for screening, adsorbing and separating inert gas materials. The method comprises the following steps: preparing an MOFs database with extensive representativeness and a potential MOFs database; the method comprises the following steps: simulating adsorption of inert gas on a widely representative MOFs database to obtain adsorption performance parameters, calculating descriptors, and establishing a data set by using the descriptors and the adsorption performance parameters; screening the potential MOFs database according to parameter range conditions such as basic LCD (Liquid Crystal Display) and PLD (Programmable Logic Device), and reducing the MOFs sample size; training a machine learning model by adopting the data set, and constructing a structure and adsorption property model; and predicting the screened MOFs with high adsorption performance on inert gas in the potential MOFs database by utilizing a structure and adsorption property model, and further applying the MOFs to experimental synthesis and performance test. According to the method, the time and the cost are greatly reduced, and the method plays a promoting role in screening high-performance materials for adsorbing and separating inert gas.

Description

Technical Field

[0001] The invention belongs to the technical field of computational chemistry and adsorption materials, and in particular relates to a method, medium and equipment for screening, adsorbing and separating inert gas materials. Background Art

[0002] Inert gases are valuable commodities in the chemical industry. Capturing these radioactive inert gases is necessary for storage, safety and environmental reasons. With the emergence of emerging nanoporous materials, such as metal organic frameworks MOF, zeolitic imidazolate frameworks ZIF, covalent organic frameworks COF and porous polymer networks PPN, the adsorption and separation technology based on new nanoporous adsorbents has a broader application prospect.

[0003] With the help of chemical theory and computer technology, a large number of nanoporous materials that may be synthesized have been constructed, including MOFs, COFs, PPNs, etc. The combination of different structural units, connection methods and topological structures can generate countless nanoporous material structures. The existence of a large number of nanoporous materials also brings many difficulties to the research and development of efficient adsorbents. It is difficult to quickly identify high-performance adsorbents through traditional trial-and-error experimental methods. Molecular simulation combined with machine learning is an effective method for quickly screening high-performance target adsorbent materials.

[0004] With the rise of artificial intelligence, more and more research is turning to machine learning (ML) assisted high-throughput computational screening (HTCS) technology to achieve the adsorption separation of inert gases. Among them, HTCS includes Grand Canonical Monte Carlo (GCMC) simulation and molecular dynamics (MD). ML mainly uses random forest (RF), extreme gradient boosting tree (XGB) and gradient boosting regression tree (GBR) for predictive calculations.

[0005] High-throughput screening can be computationally expensive. Standard computational methods such as (GCMC) simulations and classical density functional theory (CDFT) have been shown to be effective in calculating gas adsorption and separation properties, but ongoing efforts are aimed at accelerating the calculation and analysis of thermodynamic or kinetic properties relevant to industrial needs. In addition, the exponential growth of MOF structures in databases poses significant computational challenges, especially in terms of resource consumption for high-throughput screening to discover or design target materials. To address these challenges, machine learning has been widely used to predict material properties and assist in material design. Machine learning combined with high-throughput calculations of membrane separation properties has greatly accelerated the process of identifying target membrane materials. Ma et al. developed a deep neural network based on transfer learning, using SA, GSA, VF, LCD, and PLD as descriptors, and trained the DNN model with 13,506 data at 100 bar and 243 K. The R of the adsorption amount of H2 and CH4 was then predicted by transfer learning. 2 are 0.991 and 0.980, while the R 2 It is 0.092, which shows that when the base structure property descriptor is used, it is difficult to transfer the model trained with other gases to the inert gas Xe / Kr separation system. Summary of the invention

[0006] In order to overcome the problems existing in the prior art, the present invention provides a method, medium and equipment for screening, adsorbing and separating inert gas materials, which are used to overcome the current defects.

[0007] A method for screening, adsorbing and separating inert gas materials, the method comprising:

[0008] S1. Prepare a broadly representative MOFs database and a potential MOFs materials database;

[0009] S2. Computationally determine descriptors of MOFs in a broadly representative MOF database;

[0010] S3. Screening the potential MOFs material database according to the preset values ​​of the maximum free path LCD and the restricted pore PLD to obtain a screened MOFs material database;

[0011] S4. Simulating the adsorption of a single inert gas on MOFs in a widely representative MOFs database to obtain adsorption performance parameters, and establishing a data set with the adsorption performance parameters and the descriptors in S2;

[0012] S5. Divide the data set into a training set and a test set, iteratively train the machine learning model using the training set, and adjust the model's hyperparameters until the accuracy of the trade-off value of the MOFs material predicted by the machine learning model meets the requirements, use the descriptor parameters in the test set as the input of the trained machine learning model, predict the adsorption performance parameters of the test set data MOF for inert gas using the trained machine learning model, and calculate the accuracy of the predicted trade-off value of the MOFs material, compare the accuracies of different machine learning, screen out the machine learning model with the best prediction of MOF adsorption performance parameters, and construct a structure and adsorption property model;

[0013] S6. Using the structure and adsorption property model, the adsorption performance of MOSFs in the screened MOFs material database for inert gases is predicted, and high-performance MOFs that meet the requirements for adsorption and separation of inert gases are screened out.

[0014] According to the aspects described above and any possible implementation method, an implementation method is further provided, wherein the adsorption performance parameters include adsorption amount N, adsorption heat Qst, Henry coefficient K, selectivity S and trade-off value TSN; the descriptors include maximum free diameter LCD, restricted pore diameter PLD, effective pore volume APV, accessible specific surface area ASA and topological structure information.

[0015] According to the above aspects and any possible implementation, there is further provided an implementation, wherein the inert gas includes Xe, Kr and / or Ar.

[0016] According to the aspects and any possible implementations described above, an implementation is further provided, wherein the potential MOFs material database includes an eMOFs database consisting of experimentally prepared eMOFs and a hMOFs database consisting of theoretically constructed hMOFs.

[0017] According to the above aspects and any possible implementation, an implementation is further provided, wherein the descriptor adopts Zeo ++ , Poreblazer, MOFomics, ZEOMICS or TOPOS software.

[0018] According to the aspects and any possible implementation described above, an implementation is further provided, wherein in step S4, GCMC molecules are used to simulate the adsorption of a single inert gas on widely representative MOFs, specifically by inserting, deleting, moving and rotating the gas molecules in the MOFs to obtain phase equilibrium, thereby obtaining adsorption performance parameters and descriptors.

[0019] As described above and any possible implementation method, an implementation method is further provided, wherein the machine learning model includes a random forest model, a gradient boosted regression tree model, a gradient boosted regression tree model GBRT, a Bayesian regression model, a linear regression model LR, a ridge regression model Ridge, a least partial squares model PLS and a support vector machine regression model SVR.

[0020] According to the above aspects and any possible implementation, an implementation is further provided, wherein the hyperparameters are adjusted using random search or grid search.

[0021] According to the above aspects and any possible implementation, a further implementation is provided, wherein the index for evaluating accuracy in step S5 is the determination coefficient R 2 value.

[0022] According to the aspects and any possible implementations described above, an implementation is further provided, wherein high-performance adsorption MOFs are screened out in step S6, specifically comprising: screening the newly generated MOFs through the established predictive inert gas adsorption model to screen out the top 5 MOFs with the best adsorption amount.

[0023] The present invention also provides a computer-readable storage medium storing a program, wherein when the program is executed by a processor, the method described in the present invention is implemented.

[0024] The present invention also provides a computer device, comprising a processor and a memory for storing a program executable by the processor, wherein the method described is implemented when the processor executes the program stored in the memory.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The method of the present invention for organic framework adsorption materials for efficient and accurate adsorption and separation of inert gases adopts a machine learning-assisted high-throughput calculation method. It used to take three to four days to calculate a system with thousands of data, but now it only takes three days to calculate a system with tens of thousands of data. Based on Zeo++, RASPA program, structure and descriptor type obtained by thermal encoding, and adsorption performance obtained by grand canonical Monte Carlo simulation and molecular dynamics, MOFs with high adsorption performance for target gases can be quickly screened and predicted. Not only is the efficiency high, but the established machine learning model can also be applied to other MOFs, which is universal and has guiding significance for the research and development of other high-performance MOFs. Based on machine learning technology, compared with traditional experimental "trial and error" methods, grand canonical Monte Carlo simulation methods and molecular dynamics methods, the present invention is more efficient in screening high-performance MOFs, which can save a lot of time, money and human resources, and help to speed up the screening and development of high-performance MOFs materials. From a statistical perspective, the method of the present invention screens and predicts material properties. Compared with traditional experimental and simulation calculation methods, the results obtained by the method of the present invention are more reliable and universal. In the efficient screening of MOFs materials with excellent adsorption of inert gases, compared with the non-portability of traditional methods, the method of the present invention can be portable in inert gases (including Kr, Xe, Ar). BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic flow chart of the method of the present invention;

[0028] Figure 2 It is a schematic diagram of LCD and LCD / PLD data distribution of the potential MOFs material database of the present invention;

[0029] Figure 3 is a Pearson correlation coefficient graph of the first data set descriptor of the present invention;

[0030] Figure 4 It is a comparison chart of the test set and the predicted value of the inert gas selectivity value of the present invention. DETAILED DESCRIPTION

[0031] In order to better understand the technical solution of the present invention, the content of the present invention includes but is not limited to the specific implementation methods described below, and similar technologies and methods should be considered to be within the scope of protection of the present invention. In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0032] It should be clear that the embodiments described in the present invention are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] A method for screening, adsorbing and separating inert gas materials, the method comprising:

[0034] S1. Prepare a broadly representative MOFs database and a potential MOFs materials database;

[0035] S2. Computationally determine descriptors of MOFs in a broadly representative MOF database;

[0036] S3. Screening the potential MOFs material database according to the preset values ​​of the maximum free path LCD and the restricted pore PLD to obtain a screened MOFs material database;

[0037] S4. Simulating the adsorption of a single inert gas on MOFs in a widely representative MOFs database to obtain adsorption performance parameters, and establishing a data set with the adsorption performance parameters and the descriptors in S2;

[0038] S5. Divide the data set into a training set and a test set, iteratively train the machine learning model using the training set, and adjust the model's hyperparameters until the accuracy of the trade-off value of the MOFs material predicted by the machine learning model meets the requirements, use the descriptor parameters in the test set as the input of the trained machine learning model, predict the adsorption performance parameters of the test set data MOF for inert gas using the trained machine learning model, and calculate the accuracy of the predicted trade-off value of the MOFs material, compare the accuracies of different machine learning, screen out the machine learning model with the best prediction of MOF adsorption performance parameters, and construct a structure and adsorption property model;

[0039] S6. Using the structure and adsorption property model, the adsorption performance of MOSFs in the screened MOFs material database for inert gases is predicted, and high-performance MOFs that meet the requirements for adsorption and separation of inert gases are screened out.

[0040] The specific implementation process is as follows:

[0041] The present invention provides a machine learning-assisted high-throughput computing technology to efficiently and accurately screen high-performance MOFs materials for adsorption and separation of inert gases, such as Figure 1 As shown, the method comprises the following steps:

[0042] S1. Collect a widely representative MOFs database from reported literature. These MOFs have excellent adsorption and separation effects on inert gases. The present invention uses them to construct a model relationship between MOFs properties and inert gas adsorption. The potential MOFs material database includes an eMOFs database composed of experimentally prepared eMOFs and an hMOFs database composed of theoretically constructed hMOFs. The combination of different structural units, connection methods and topological structures can generate countless MOFs structures. The MOFs database prepared by researchers using experiments and synthesized with the help of chemical theory and computer technology is a potential MOFs material database.

[0043] The inert gas is Xe, Kr or Ar.

[0044] S2. The largest free diameter LCD (LCD), limited pore diameter PLD (PLD), available pore volume APV (Available Pore Volume), accessible surface area ASA (Accessible Surface Area, ASA) and topological structure information of MOFs in a widely representative MOFs database are calculated using Zeo++, Poreblazer, MOFomics or ZEOMICS and TOPOS as descriptors. These descriptors are used as elements in the following data sets;

[0045] S3, for the specific inert gas type to be adsorbed and separated, find the basic LCD, PLD and other parameter ranges of MOFs suitable for the separation of such inert gas, that is, set the LCD and PLD setting value parameters or setting value range parameters of MOFs suitable for the separation of a certain inert gas in advance, sample these pre-set values ​​to screen the potential MOFs material database, screen out a part of MOFs materials whose LCD and PLD of the potential MOFs materials are within the set value parameters or the set value range parameters, obtain the screened potential MOFs material database, and process the screened potential MOFs material database in the subsequent steps, thereby reducing the amount of calculation;

[0046] S4, using the GCMC method to simulate the adsorption of a single inert gas on MOFs in a widely representative MOFs database, and obtain the adsorption performance parameters of MOFs for the inert gas in the system, including (adsorption amount N, adsorption heat Qst and Henry coefficient K, selectivity S and trade-off value TSN), and combining the adsorption performance parameters obtained in this step with the descriptors obtained in S2 to establish a data set;

[0047] S5. Divide the data set into a training set and a test set, iteratively train the machine learning model using the training set, and adjust the model's hyperparameters until the accuracy of the trade-off value of the MOFs material predicted by the machine learning model meets the requirements, use the descriptor parameters in the test set as the input of the trained machine learning model, use the trained machine learning model to predict the adsorption performance parameters of the MOF for inert gas corresponding to the descriptor parameters of the test set, and calculate the accuracy of the predicted trade-off value of the MOFs material, compare the accuracies of different machine learning, screen out the best machine learning model for predicting the MOFs adsorption performance parameters, and construct a structure and adsorption property model;

[0048] Among them, machine learning models include gradient boosted regression tree GBRT, Bayesian regression model, linear regression model LR, ridge regression model Ridge, least partial squares model PLS, support vector machine regression model SVR, and random forest model. Adjustment of model hyperparameters includes but is not limited to random search and grid search.

[0049] Among them, the indicator for evaluating accuracy is the determination coefficient R 2 value.

[0050] S6. Use the structure and adsorption property model to predict the adsorption performance of the potential MOFs material database for inert gases after screening, and screen out high-performance MOFs that meet the requirements for adsorption and separation of inert gases.

[0051] This step also specifically includes: predicting and screening out the top five MOFs with the best adsorption capacity from the screened potential MOFs material database through the established structure and adsorption property model. The top five MOFs materials have the best performance in adsorption capacity and are the most advantageous materials in potential applications.

[0052] Preferably, the present invention further comprises step S7. prediction and accuracy evaluation of the adsorption performance of MOFs for inert gas: based on the prediction model and prediction situation obtained in step S6, the prediction results are analyzed to evaluate the accuracy of the results predicted by the prediction model used. To better describe the present invention, a specific example is given below.

[0053] (1) Building a database:

[0054] Establish a broadly representative MOFs database: By collecting about 900 MOFs materials with excellent adsorption and separation properties for inert gases from the literature, a broadly representative MOFs database is constructed.

[0055] Establishment of a database of potential MOFs:

[0056] 9146 data in the CoRE MOF 2019 dataset were selected as the potential MOFs database. The database already contains LCD and PLD data, which can be directly analyzed. Literature search revealed that the LCD of MOFs suitable for the separation of Xe and Kr mixed gases is 4 to 8, and the LCD / PLD ratio is between 1 and 2. These MOFs have relatively high selectivity for Xe / Kr. Figure 2 As shown in the figure, the distribution of LCD and LCD / PLD in the CoREMOF 2019 dataset is analyzed, and the database will be screened later.

[0057] (2) Determine the descriptor:

[0058] The MOFs descriptors in a widely representative MOFs database were determined, and the geometric and energy descriptors of the MOFs were calculated using Zeo++ and RASPA programs, including information such as the maximum free diameter LCD, confined pore diameter PLD, effective pore volume APV, accessible specific surface area ASA, and topological structure.

[0059] (3) The GCMC method is used to simulate the adsorption of Xe or Kr on MOFs in a widely representative MOFs database to obtain the adsorption performance parameters of MOFs for inert gases in the system, including the adsorption amount N, adsorption heat Qst and Henry coefficient K, selectivity S and trade-off value TSN. The adsorption performance parameters and descriptors in this step are used to establish a data set.

[0060] like Figure 3 As shown in the figure, the Pearson Correlation Coefficients correlation coefficients of the descriptors for predicting adsorption selectivity are finally determined. It can be seen that the different structural property descriptors of MOFs in the widely representative MOFs database have a significant impact on S Xe / Kr The greater the correlation, the higher the prediction accuracy. These and other descriptors can be used for subsequent model building.

[0061] (4) Using pre-set LCD and PLD setting value parameters or parameter ranges to screen the potential MOFs database, MOFs that meet the screening requirements are combined into a new database, i.e., the screened potential MOFs database.

[0062] Preferably, in this step, the potential MOFs database may include data of LCD and PLD, and the potential MOFs database may be directly analyzed and screened, and a portion of MOFs may be deleted to obtain a screened potential MOFs database, thereby reducing the amount of calculation in subsequent operations.

[0063] (5) Divide the training set and test set:

[0064] Based on the data set established in step (3), the data set is randomly divided into a training set train set and a test set test set in a ratio of 8:2.

[0065] (6) Establish a descriptor-performance prediction model:

[0066] Select the random forest model RF, gradient boosted regression tree GBRT, Bayesian regression model, linear regression model LR, ridge regression model Ridge and support vector machine regression model SVR as the models to be trained;

[0067] Use training set and test set and match the determination coefficient R 2 Value, quantitatively describe the accuracy of the prediction results of the machine learning model, select R 2 The algorithm with the highest value is taken as the best algorithm, and this model is selected as the best prediction model. The prediction model is used to predict the potential MOFs database after screening obtained in step (4) to obtain high-performance MOFs that meet the requirements for adsorption and separation of inert gas materials.

[0068] (7) Prediction and accuracy evaluation of MOFs adsorption performance for inert gases:

[0069] Based on the prediction model and prediction situation obtained in step (6), the prediction results are analyzed. Figure 4 As shown in the figure, the Bayesian regression model, linear regression model LR, ridge regression model Ridge, least partial squares model PLS and support vector machine regression model SVR can all be used for prediction, and the prediction results can meet the expected requirements; for the adsorption selectivity of inert gases, the prediction results of random forest model RF and gradient boosted regression tree GBRT are more accurate, and the test set and training set R 2 Reaching above 0.90. The random forest model RF has relatively good prediction results for the adsorption selectivity of inert gases. This is because the random forest model RF uses an unbiased estimate for the generalization error and has a strong model generalization ability.

[0070] The method provided in the above embodiment of the present invention can quickly screen and predict MOFs with high adsorption performance for the target gas. It is not only highly efficient, but the established machine learning model can also be used to predict other MOFs data set materials, which has guiding significance. Compared with the traditional experimental "trial and error" method, grand canonical Monte Carlo simulation method and molecular dynamics method, the machine learning model established by the machine learning prediction and screening method of the present invention is more efficient in screening high-performance MOFs than the classical simulation method, which can save money and human resources, and is conducive to the screening and research and development process of high-performance MOFs materials.

[0071] As an embodiment disclosed in the present invention, the present invention further provides a computer-readable storage medium storing a program, and when the program is executed by a processor, the method described is implemented.

[0072] As an embodiment disclosed in the present invention, the present invention further provides a device, including a processor and a memory for storing a program executable by the processor, and when the processor executes the program stored in the memory, the described method is implemented.

[0073] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0074] The above description shows and describes several preferred embodiments of the present invention, but as mentioned above, it should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the application concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not depart from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for screening, adsorbing and separating inert gas materials, characterized in that: The method comprises: S1. Prepare a broadly representative MOFs database and a potential MOFs materials database; S2. Computationally determine descriptors of MOFs in a broadly representative MOF database; S3. Screening the potential MOFs material database according to the preset values ​​of the maximum free path LCD and the restricted pore PLD to obtain a screened MOFs material database; S4. Simulating the adsorption of a single inert gas on MOFs in a widely representative MOFs database to obtain adsorption performance parameters, and establishing a data set with the adsorption performance parameters and the descriptors in S2; S5. Divide the data set into a training set and a test set, iteratively train the machine learning model using the training set, and adjust the model's hyperparameters until the accuracy of the trade-off value of the MOFs material predicted by the machine learning model meets the requirements, use the descriptor parameters in the test set as the input of the trained machine learning model, predict the adsorption performance parameters of the test set data MOF for inert gas using the trained machine learning model, and calculate the accuracy of the predicted trade-off value of the MOFs material, compare the accuracies of different machine learning, screen out the machine learning model with the best prediction of MOF adsorption performance parameters, and construct a structure and adsorption property model; S6. Using the structure and adsorption property model, the adsorption performance of MOSFs in the screened MOFs material database for inert gases is predicted, and high-performance MOFs that meet the requirements for adsorption and separation of inert gases are screened out.

2. The method according to claim 1, characterized in that The adsorption performance parameters include adsorption amount N, adsorption heat Qst, Henry coefficient K, selectivity S and trade-off value TSN; the descriptors include maximum free diameter LCD, restricted pore diameter PLD, effective pore volume APV, accessible specific surface area ASA and topological structure information.

3. The method according to claim 1, characterized in that The inert gas is Xe, Kr and / or Ar.

4. The method according to claim 1, characterized in that The potential MOFs material database includes an eMOFs database consisting of experimentally prepared eMOFs and a hMOFs database consisting of theoretically constructed hMOFs.

5. The method according to claim 2, characterized in that: The descriptors were calculated using Zeo++, Poreblazer, MOFomics, ZEOMICS or TOPOS software.

6. The method according to claim 2, characterized in that In step S4, GCMC molecules are used to simulate the adsorption of a single inert gas on widely representative MOFs, specifically by inserting, deleting, moving and rotating the gas molecules in the MOFs to obtain phase equilibrium, thereby obtaining adsorption performance parameters and descriptors.

7. The method according to claim 6, characterized in that The machine learning models include random forest model, gradient boosted regression tree model, gradient boosted regression tree model GBRT, Bayesian regression model, linear regression model LR, ridge regression model Ridge, least partial squares model PLS and support vector machine regression model SVR.

8. The method according to claim 1, characterized in that: The hyperparameters are adjusted by random search or grid search.

9. A computer-readable storage medium storing a program, characterized in that: When the program is executed by a processor, the method described in any one of claims 1 to 8 is implemented.

10. A computer device comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, the method described in any one of claims 1 to 8 is implemented.

Images