Machine learning assisted polyethylene reaction performance prediction model training method, prediction method and device

Abstract

The invention discloses a machine learning assisted polyethylene reaction performance prediction model training method, prediction method and device. The method comprises the following steps: acquiring a training set; screening feature items used for model training; obtaining a gradient boosting regression model for catalytic activity, a gradient boosting regression model for molecular weight and a gradient boosting regression model for molecular weight distribution; extracting feature items for model training from the data of the training set so as to obtain feature vectors; and respectively inputting the feature vectors into each model so as to train each model, thereby respectively obtaining hyper-parameters of the trained gradient-boosted regression model for catalytic activity, hyper-parameters of the trained gradient-boosted regression model for molecular weight and hyper-parameters of the trained gradient-boosted regression model for molecular weight distribution. According to the method, a model relationship between input characteristics and polymerization results (including catalytic activity, molecular weight, molecular weight distribution and the like) is established through training set learning.

Description

Technical Field

[0001] This application relates to the field of polyethylene technology, specifically to a machine learning-assisted method for training a polyethylene reaction performance prediction model, a machine learning-assisted device for training a polyethylene reaction performance prediction model, and a method for predicting polyethylene reaction performance. Background Technology

[0002] Polyethylene is a polymer material produced by the polymerization of ethylene monomers. It has wide applications in chemical engineering, aerospace, marine engineering, and biomedicine. The chemical structure, molecular weight, and molecular weight distribution of polyethylene are important quality indicators, closely related to the material's physical and processing properties. These properties largely depend on the polymerization conditions, including the selection of catalysts and reaction conditions such as temperature and pressure. Appropriate selection of these conditions can improve reaction efficiency and product quality.

[0003] Currently, polyolefin catalysts mainly include Ziegler-Natta catalysts, metallocene catalysts, and non-metallocene catalysts. Among them, Ziegler-Natta catalysts have multiple active sites and complex catalytic mechanisms. Compared with traditional Ziegler-Natta catalysts, single-active catalytic systems have a single active site and are characterized by narrow molecular weight distribution and uniform distribution of comonomers. Compared with the ligands of metallocene catalysts, non-metallocene ligands have relatively low synthesis costs and are easy to modify, offering the advantage of ligand diversity. FI catalysts are a class of pre-transition metal catalysts developed by the Fujita team at Mitsui Chemicals, using phenoxyimine compounds as ligands. Under mild experimental conditions, FI catalysts exhibit very high activity for ethylene polymerization and show promising development prospects.

[0004] To improve the performance of polyethylene, researchers are continuously developing polyethylene products with different molecular weights and molecular weight distributions to meet the needs of various fields for polyethylene resins. This process involves extensive catalyst screening and condition optimization, resulting in significant time costs and resource waste, thus posing challenges to material research and development. Summary of the Invention

[0005] The purpose of this invention is to provide a machine learning-assisted method for training a polyethylene reaction performance prediction model to at least solve one of the above-mentioned technical problems.

[0006] One aspect of the present invention provides a machine learning-assisted method for training a polyethylene reactivity prediction model, the machine learning-assisted method for training a polyethylene reactivity prediction model includes:

[0007] Obtain the training set;

[0008] Filter the feature terms used for model training;

[0009] Gradient-enhanced regression models were used to obtain catalytic activity, molecular weight, and molecular weight distribution.

[0010] Feature vectors are obtained by extracting feature terms from the training set data for model training.

[0011] The feature vectors are input into the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution, respectively, to train each model, thereby obtaining the hyperparameters of the trained gradient boosting regression model for catalytic activity, the trained gradient boosting regression model for molecular weight, and the trained gradient boosting regression model for molecular weight distribution, respectively.

[0012] Optionally, the training set is obtained in the following manner:

[0013] A database for ethylene homopolymerization was established, containing three central metal atoms (Ti, Zr, Hf), 32 R1 ligands, 14 R2 ligands, and 10 R3 ligands, as well as catalytic activity, molecular weight, and molecular weight distribution measured under different operating conditions.

[0014] R1 ligands include: phenyl, methyl, 2-isopropylphenyl, cyclopropyl, cyclopentyl, ethyl, propyl, isopropyl, n-hexane, cyclobutyl, 2-methylcyclohexyl, cyclohexyl, 2,3,4,5,6-pentafluorophenyl, 2-methylphenyl, 2-tert-butylphenyl, 3,5-di-tert-butylphenyl, 4-tert-butylphenyl, cyclooctyl, cycloheptyl, 2-phenylethyl, 3-phenylpropyl, 2,4,6-trifluorophenyl, 2,6-difluorophenyl, 2-fluorophenyl, 3,4,5-trifluorophenyl, 3,5-difluorophenyl, 4-fluorophenyl, 4-(4-vinylphenyl)-2,6-difluorophenyl, 4-(4-ethylphenyl)-2,6-difluorophenyl, 2,6-dimethylphenyl, 2,6-diisopropylphenyl, naphthyl;

[0015] R2 ligands include: tert-butyl, adamantane, isopropyl, cumyl, methyl, phenyl, trimethylsilyl, cyclopentyl, hydrogen atom, fluorine atom, chlorine atom, bromine atom, and iodine atom;

[0016] R3 ligands include: methyl, phenyl, tert-butyl, cumyl, methoxy, hydrogen, fluorine, bromine, and iodine.

[0017] Optionally, the selection of feature terms for model training includes:

[0018] SHAP analysis was introduced to explore the correlation between various characteristic quantities and prediction results, as well as the importance of the characteristics. The 10 characteristics that contributed the most to the prediction of catalytic activity, molecular weight, and molecular weight distribution were screened out, including the central metal (M), R1 ligand (R1), R2 ligand (R2), R3 ligand (R3), polymerization temperature (T), polymerization pressure (P), Al / M ratio, polymerization time (Time), catalyst dosage (Cat), and cocatalyst dosage (Cocat).

[0019] Optionally, the step of inputting the feature vector into the gradient boosting regression model to train the model and thereby obtaining the hyperparameters of the trained gradient boosting regression model includes:

[0020] During training, ten-fold cross-validation and Bayesian optimization are used to determine the optimal combination of hyperparameters.

[0021] Optionally, regarding the catalytic activity prediction: After Bayesian optimization, we finally determined the hyperparameters of the catalytic activity prediction model: learning rate: 0.37798365974066117, maximum depth: 5, number of base learners: 77, and random seed for the tree estimator: 23. The coefficient of determination is 0.9963, and the mean absolute error is only 3.7090 kg PE / (mmol cat·h).

[0022] Optionally, regarding the molecular weight prediction: After Bayesian optimization, we finally determined the hyperparameters of the molecular weight prediction model: learning rate: 0.3046923124838408, maximum depth: 5, number of base learners: 74, and random seed for the tree estimator: 27. The coefficient of determination is 0.9873, and the mean absolute error is only 2.8007 × 10⁴ g / mol.

[0023] Optionally, for molecular weight distribution prediction: After Bayesian optimization, we finally determined the hyperparameters of the molecular weight distribution prediction model: learning rate: 0.38109249044055854, maximum depth: 5, number of base learners: 77, and random seed for the tree estimator: 23. The coefficient of determination is 0.9609, and the mean absolute error is only 0.1342.

[0024] This application also provides a machine learning-assisted polyethylene reactivity prediction model training device, the machine learning-assisted polyethylene reactivity prediction model training device comprising:

[0025] A training set acquisition module, which is used to acquire a training set;

[0026] A filtering module, which is used to filter feature terms for model training;

[0027] The model acquisition module is used to acquire gradient boosting regression models for catalytic activity, molecular weight, and molecular weight distribution.

[0028] The feature extraction module is used to extract feature terms from the training set data for model training to obtain feature vectors;

[0029] The training module is used to input feature vectors into the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution, respectively, to train each model, thereby obtaining the hyperparameters of the trained gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution, respectively.

[0030] This application also provides a method for predicting the reactivity of polyethylene, the method comprising:

[0031] Obtain the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution trained by the machine learning-assisted polyethylene reaction performance prediction model training method described above;

[0032] The catalytic activity is coupled using a gradient boosting regression model, the molecular weight is coupled using a gradient boosting regression model, and the molecular weight distribution is coupled using a gradient boosting regression model to obtain a coupled model;

[0033] Obtain the features to be predicted;

[0034] The features to be predicted are input into the coupling model to obtain catalytic activity information, molecular weight information, and molecular weight distribution information.

[0035] Beneficial effects

[0036] The machine learning-assisted polyethylene reactivity prediction model training method of this application has the following advantages:

[0037] 1. This application establishes a model relationship between input features and polymerization results (including catalytic activity, molecular weight, molecular weight distribution, etc.) through training set learning. Users can directly obtain the predicted results of catalytic activity, molecular weight, and molecular weight distribution by inputting feature variables;

[0038] 2. It couples three machine learning models: catalytic activity prediction, molecular weight prediction, and molecular weight distribution prediction. Users can directly obtain the predicted results of catalytic activity, molecular weight, and molecular weight distribution by inputting the central transition metal, ligand R1, ligand R2, ligand R3, temperature, pressure, polymerization time, catalyst dosage, co-catalyst dosage, and co-catalyst-catalyst concentration ratio. Attached Figure Description

[0039] Figure 1 This is a schematic flowchart of a machine learning-assisted polyethylene reactivity prediction model training method according to an embodiment of this application.

[0040] Figure 2 It is used to implement Figure 1 The diagram shows an electronic device for training a machine learning-assisted polyethylene reaction performance prediction model.

[0041] Figure 3 This is a schematic diagram of the structure of a FI catalyst according to an embodiment of this application.

[0042] Figure 4 This is a scatter plot of catalytic activity prediction from one embodiment of this application.

[0043] Figure 5 This is a scatter plot of molecular weight information prediction from one embodiment of this application.

[0044] Figure 6 This is a scatter plot of molecular weight distribution information prediction from one embodiment of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0046] like Figure 1 The machine learning-assisted training method for predicting the reactivity of polyethylene, as shown, includes:

[0047] Obtain the training set;

[0048] Filter the feature terms used for model training;

[0049] Gradient-enhanced regression models were used to obtain catalytic activity, molecular weight, and molecular weight distribution.

[0050] Feature vectors are obtained by extracting feature terms from the training set data for model training.

[0051] The feature vectors are input into the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution, respectively, to train each model, thereby obtaining the hyperparameters of the trained gradient boosting regression model for catalytic activity, the trained gradient boosting regression model for molecular weight, and the trained gradient boosting regression model for molecular weight distribution, respectively.

[0052] In this embodiment, the training set is obtained in the following manner:

[0053] A database for ethylene homopolymerization was established, containing three central metal atoms (Ti, Zr, Hf), 32 R1 ligands, 14 R2 ligands, and 10 R3 ligands, as well as catalytic activity, molecular weight, and molecular weight distribution measured under different operating conditions.

[0054] R1 ligands include: phenyl, methyl, 2-isopropylphenyl, cyclopropyl, cyclopentyl, ethyl, propyl, isopropyl, n-hexane, cyclobutyl, 2-methylcyclohexyl, cyclohexyl, 2,3,4,5,6-pentafluorophenyl, 2-methylphenyl, 2-tert-butylphenyl, 3,5-di-tert-butylphenyl, 4-tert-butylphenyl, cyclooctyl, cycloheptyl, 2-phenylethyl, 3-phenylpropyl, 2,4,6-trifluorophenyl, 2,6-difluorophenyl, 2-fluorophenyl, 3,4,5-trifluorophenyl, 3,5-difluorophenyl, 4-fluorophenyl, 4-(4-vinylphenyl)-2,6-difluorophenyl, 4-(4-ethylphenyl)-2,6-difluorophenyl, 2,6-dimethylphenyl, 2,6-diisopropylphenyl, naphthyl;

[0055] R2 ligands include: tert-butyl, adamantane, isopropyl, cumyl, methyl, phenyl, trimethylsilyl, cyclopentyl, hydrogen atom, fluorine atom, chlorine atom, bromine atom, and iodine atom;

[0056] R3 ligands include: methyl, phenyl, tert-butyl, cumyl, methoxy, hydrogen, fluorine, bromine, and iodine.

[0057] In this embodiment, the selection of feature terms for model training includes:

[0058] SHAP analysis was introduced to explore the correlation between various characteristic quantities and prediction results, as well as the importance of the characteristics. The 10 characteristics that contributed the most to the prediction of catalytic activity, molecular weight, and molecular weight distribution were screened out, including the central metal (M), R1 ligand (R1), R2 ligand (R2), R3 ligand (R3), polymerization temperature (T), polymerization pressure (P), Al / M ratio, polymerization time (Time), catalyst dosage (Cat), and cocatalyst dosage (Cocat).

[0059] In this embodiment, the step of inputting the feature vector into the gradient boosting regression model to train the model and thereby obtaining the hyperparameters of the trained gradient boosting regression model includes:

[0060] During training, ten-fold cross-validation and Bayesian optimization are used to determine the optimal combination of hyperparameters.

[0061] See Figure 4 In this embodiment, the catalytic activity prediction model, after Bayesian optimization, has the following hyperparameters: learning rate: 0.37798365974066117, maximum depth: 5, number of base learners: 77, tree estimator random seed: 23. The coefficient of determination is 0.9963, and the mean absolute error is only 3.7090 kg PE / (mmol cat·h).

[0062] See Figure 5 In this embodiment, the molecular weight prediction, after Bayesian optimization, finally determined the hyperparameters of the molecular weight prediction model: learning rate: 0.3046923124838408, maximum depth: 5, number of base learners: 74, random seed for tree estimator: 27. The coefficient of determination is 0.9873, and the mean absolute error is only 2.8007 × 10⁴ g / mol.

[0063] See Figure 6 In this embodiment, for molecular weight distribution prediction: after Bayesian optimization, we finally determined the hyperparameters of the molecular weight distribution prediction model: learning rate: 0.38109249044055854, maximum depth: 5, number of base learners: 77, and random seed for the tree estimator: 23. The coefficient of determination is 0.9609, and the mean absolute error is only 0.1342.

[0064] This application also provides a machine learning-assisted polyethylene reactivity prediction model training device, the machine learning-assisted polyethylene reactivity prediction model training device comprising:

[0065] A training set acquisition module, which is used to acquire a training set;

[0066] A filtering module, which is used to filter feature terms for model training;

[0067] The model acquisition module is used to acquire gradient boosting regression models for catalytic activity, molecular weight, and molecular weight distribution.

[0068] The feature extraction module is used to extract feature terms from the training set data for model training to obtain feature vectors;

[0069] The training module is used to input feature vectors into the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution, respectively, to train each model, thereby obtaining the hyperparameters of the trained gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution, respectively.

[0070] This application also provides a method for predicting the reactivity of polyethylene, the method comprising:

[0071] Obtain the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution trained by the machine learning-assisted polyethylene reaction performance prediction model training method described above;

[0072] The catalytic activity is coupled using a gradient boosting regression model, the molecular weight is coupled using a gradient boosting regression model, and the molecular weight distribution is coupled using a gradient boosting regression model to obtain a coupled model;

[0073] Obtain the features to be predicted;

[0074] The features to be predicted are input into the coupling model to obtain catalytic activity information, molecular weight information, and molecular weight distribution information.

[0075] In this embodiment, the various models of this application are coupled together through web programming. Specifically, a Flask application folder was created, and HTML, CSS, and Python backend code were written to ultimately implement the web application. Users can directly obtain predicted results for catalytic activity, molecular weight, and molecular weight distribution by inputting features such as the central transition metal, ligand R1, ligand R2, ligand R3, temperature, pressure, polymerization time, catalyst dosage, co-catalyst dosage, and co-catalyst-to-catalyst concentration ratio. Python can be trained by importing prediction functions from the training model module, and one program outputs three results, which is equivalent to connecting three models together and outputting the results together.

[0076] The following examples further illustrate this application in detail. It is understood that these examples do not constitute any limitation on this application.

[0077] A method for constructing and applying a machine learning-assisted prediction model for the reaction performance of polyethylene includes the following steps:

[0078] Data Collection: A database for ethylene homopolymerization was established, containing three central metal atoms (Ti, Zr, Hf), 32 R1 ligands, 13 R2 ligands, and 10 R3 ligands, as well as catalytic activity, molecular weight, and molecular weight distribution measured under different operating conditions. Specifically, the operating conditions of this application may include the following: temperature (°C), pressure (MPa), co-catalyst / catalyst ratio, polymerization time (min), catalyst dosage (μmol), and co-catalyst dosage (mmol). In this embodiment, activity information (kg PE / mmol cat·h), molecular weight information (g / mol), and molecular weight distribution information for each combination were obtained using the above operating conditions.

[0079] Data preprocessing: The catalyst structure, reaction conditions, and polymerization results were compiled into a database, and missing values ​​in the original dataset were removed. The constructed dataset was further divided into a training dataset (80%) and a test dataset (20%) according to the principle of independent and identically distributed data, which will be used as sample data for training machine learning models.

[0080] Feature Selection: SHAP analysis was introduced to explore the correlation and importance of each feature quantity and prediction result, selecting the 10 features that contribute most to the prediction of catalytic activity, molecular weight, and molecular weight distribution. These features include the central metal (M) of the FI catalyst, R1 ligand (R1), R2 ligand (R2), R3 ligand (R3), polymerization temperature (T), polymerization pressure (P), Al / M ratio, polymerization time (Time), catalyst dosage (Cat), and cocatalyst dosage (Cocat). In this embodiment, SHAP is a game theory-based method that can be used to reveal the complex relationship between input features and output targets, to explain the machine learning model in detail, and to quantitatively understand the impact of input features on the best model prediction. Basic Idea: Calculate the marginal contribution of a feature when added to the model, and then, considering the different marginal contributions of that feature across all feature sequences, take the average, i.e., the SHAP value of that feature.

[0081] Model Training: During training, 10-fold cross-validation and Bayesian optimization are used to determine the optimal combination of hyperparameters. After the machine learning model is trained, the next step is to evaluate its performance. For objective and accurate evaluation, the test samples should be representative and unrelated to the training samples. Otherwise, false test scores may be given. To avoid overfitting, a common approach in supervised machine learning experiments is to reserve a portion of the existing dataset as a test set. However, dividing the entire dataset into three sets—training, testing, and validation—significantly reduces the amount of data available for modeling. Therefore, we apply cross-validation, specifically 10-fold cross-validation, dividing the training dataset into 10 subsets. In 10 training iterations, 9 subsets are used as training data each time, and the remaining subset is used as validation data. This eliminates the need for a separate validation set, and the test set is always used for the final evaluation of the model. Every data point has the opportunity to serve as validation data, further ensuring the robustness and generalization ability of the model.

[0082] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and will not be repeated here.

[0083] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements the above-described machine learning-assisted polyethylene reaction performance prediction model training method.

[0084] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, enables the above-described machine learning-assisted method for training a polyethylene reaction performance prediction model.

[0085] Figure 2 This is an exemplary structural diagram of an electronic device capable of implementing a machine learning-assisted polyethylene reaction performance prediction model training method according to an embodiment of this application.

[0086] like Figure 2 As shown, the electronic device includes an input device 501, an input interface 502, a central processing unit 503, a memory 504, an output interface 505, and an output device 506. The input interface 502, central processing unit 503, memory 504, and output interface 505 are interconnected via a bus 507. The input device 501 and output device 506 are connected to the bus 507 via the input interface 502 and output interface 505, respectively, and thus connected to other components of the electronic device. Specifically, the input device 501 receives input information from the outside and transmits it to the central processing unit 503 via the input interface 502. The central processing unit 503 processes the input information based on computer-executable instructions stored in the memory 504 to generate output information, temporarily or permanently storing the output information in the memory 504, and then transmitting the output information to the output device 506 via the output interface 505. The output device 506 outputs the output information to the outside of the electronic device for user use.

[0087] In other words, Figure 2 The illustrated electronic device may also be implemented as including: a memory storing computer-executable instructions; and one or more processors, which can be coupled when executing the computer-executable instructions. Figure 1 The method described is a machine learning-assisted training method for predicting the reactivity of polyethylene.

[0088] In one embodiment, Figure 2 The electronic device shown can be implemented as including: a memory 504 configured to store executable program code; and one or more processors configured to run the executable program code stored in the memory 504 to execute the machine learning-assisted polyethylene reaction performance prediction model training method in the above embodiments.

[0089] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0090] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0091] Computer-readable media include both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, DVD or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0092] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which includes one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutively marked blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or the overall flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0093] In this embodiment, the processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0094] Memory can be used to store computer programs and / or modules. The processor implements various functions of the device / terminal equipment by running or executing the computer programs and / or modules stored in the memory, and by accessing data stored in the memory. Memory can mainly include a program storage area and a data storage area. The program storage area can store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area can store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). In addition, memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0095] In this embodiment, if the modules / units integrated into the device / terminal equipment are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

[0096] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0097] Furthermore, it is clear that the word "comprising" does not exclude other units or steps. Multiple units, modules, or devices recited in the apparatus claims may also be implemented by a single unit or overall apparatus via software or hardware.

[0098] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A machine learning assisted polyethylene reaction performance prediction model training method, characterized in that, The machine learning assisted polyethylene reaction performance prediction model training method comprises: obtaining a training set; screening feature items for model training; obtaining a gradient boosting regression model for catalytic activity, a gradient boosting regression model for molecular weight, and a gradient boosting regression model for molecular weight distribution; extracting feature items for model training from the data of the training set to obtain a feature vector; inputting the feature vector into the gradient boosting regression model for catalytic activity, the gradient boosting regression model for molecular weight, and the gradient boosting regression model for molecular weight distribution respectively to train each model, thereby obtaining the hyperparameters of the trained gradient boosting regression model for catalytic activity, the hyperparameters of the trained gradient boosting regression model for molecular weight, and the hyperparameters of the trained gradient boosting regression model for molecular weight distribution respectively.

2. The machine learning assisted polyethylene reaction performance prediction model training method of claim 1, wherein, The training set is obtained by the following method: An ethylene homopolymerization reaction database is established, which contains 3 central metal atoms (Ti, Zr, Hf), 32 R1 ligands, 14 R2 ligands, and 10 R3 ligands, as well as measured catalytic activity, molecular weight, and molecular weight distribution under different operating conditions; wherein, R1 ligands include: phenyl, methyl, 2-isopropylphenyl, cyclopropyl, cyclopentyl, ethyl, propyl, isopropyl, n-hexane, cyclobutyl, 2-methylcyclohexyl, cyclohexyl, 2,3,4,5,6-pentafluorophenyl, 2-methylphenyl, 2-tert-butylphenyl, 3,5-di-tert-butylphenyl, 4-tert-butylphenyl, cyclooctyl, cycloheptyl, 2-phenethyl, 3-phenylpropyl, 2,4,6-trifluorophenyl, 2,6-difluorophenyl, 2-fluorophenyl, 3,4,5-trifluorophenyl, 3,5-difluorophenyl, 4-fluorophenyl, 4-(4-vinylphenyl)-2,6-difluorophenyl, 4-(4-ethylphenyl)-2,6-difluorophenyl, 2,6-dimethylphenyl, 2,6-diisopropylphenyl, naphthyl; R2 ligands include: tert-butyl, adamantyl, isopropyl, cumyl, methyl, phenyl, trimethylsilyl, cyclopentyl, hydrogen atom, fluorine atom, chlorine atom, bromine atom, iodine atom; R3 ligands include: methyl, phenyl, tert-butyl, cumyl, methoxy, hydrogen atom, fluorine atom, bromine atom, iodine atom.

3. The machine learning assisted polyethylene reaction performance prediction model training method of claim 2, wherein, The screening of feature items for model training comprises: Introducing SHAP analysis to deeply explore the correlation between each feature quantity and the prediction result and the feature importance, and screening out the top 10 features that contribute most to the prediction of catalytic activity, molecular weight, and molecular weight distribution, including the central metal (M) of the FI catalyst, the R1 ligand (R1), the R2 ligand (R2), the R3 ligand (R3), the polymerization temperature (T), the polymerization pressure (P), the Al / M ratio, the polymerization time (Time), the catalyst dosage (Cat), and the cocatalyst dosage (Cocat).

4. The machine learning assisted polyethylene reaction performance prediction model training method of claim 3, wherein, The inputting of the feature vector into the gradient boosting regression model for training the model, thereby obtaining the hyperparameters of the trained gradient boosting regression model, comprises: During the training process, the ten-fold cross-validation method and Bayesian optimization are used to determine the optimal combination of hyperparameters.

5. The machine learning assisted polyethylene reaction performance prediction model training method of claim 4, wherein, The catalytic activity prediction: through Bayesian optimization, we finally determine the hyperparameters of the catalytic activity prediction model, learning rate: 0.37798365974066117, maximum depth: 5, base learner number: 77, tree estimator random seed: 23, coefficient of determination 0.9963, mean absolute error only 3.7090 kg PE / (mmol cat·h).

6. The machine learning assisted polyethylene reaction performance prediction model training method of claim 5, wherein, The molecular weight prediction: After Bayesian optimization, we finally determined the hyperparameters of the molecular weight prediction model, learning rate: 0.3046923124838408, maximum depth: 5, base learner number: 74, tree estimator random seed: 27, coefficient of determination 0.9873, mean absolute error only 2.8007 x 10 4 g / mol.

7. The machine learning assisted polyethylene reaction performance prediction model training method of claim 6, wherein, Molecular weight distribution prediction: through Bayesian optimization, we finally determine the hyperparameters of the molecular weight distribution prediction model, learning rate: 0.38109249044055854, maximum depth: 5, base learner number: 77, tree estimator random seed: 23, coefficient of determination 0.9609, mean absolute error only 0.1342. 8.A machine learning-assisted polyethylene reaction performance prediction model training apparatus, characterized in that, The machine learning assisted polyethylene reaction performance prediction model training device comprises: A training set acquisition module for acquiring a training set; A screening module for screening feature items for model training; A model acquisition module for acquiring a catalytic activity gradient boosting regression model, a molecular weight gradient boosting regression model, and a molecular weight distribution gradient boosting regression model; A feature extraction module for extracting feature items for model training from the data of the training set to obtain a feature vector; A training module for inputting the feature vector into the catalytic activity gradient boosting regression model, the molecular weight gradient boosting regression model, and the molecular weight distribution gradient boosting regression model to train each model, thereby obtaining the hyperparameters of the trained catalytic activity gradient boosting regression model, the hyperparameters of the trained molecular weight gradient boosting regression model, and the hyperparameters of the trained molecular weight distribution gradient boosting regression model.

9. A method of predicting the reactivity of polyethylene, characterized by, The polyethylene reaction performance prediction method comprises: Acquiring the catalytic activity gradient boosting regression model, the molecular weight gradient boosting regression model, and the molecular weight distribution gradient boosting regression model trained by the machine learning assisted polyethylene reaction performance prediction model training method according to any one of claims 1 to 7; Coupling the catalytic activity gradient boosting regression model, the molecular weight gradient boosting regression model, and the molecular weight distribution gradient boosting regression model to obtain a coupled model; Obtaining the to-be-predicted features; Inputting the to-be-predicted features into the coupled model to obtain catalytic activity information, molecular weight information, and molecular weight distribution information.

Images