Multi-target product process regulation and control method and system based on machine learning

Abstract

The invention provides a multi-target product process regulation and control method and system based on machine learning, and relates to the technical field of machine learning. According to the regulation and control method, an original data set is classified into three classes of MG, EG and ET through a yield result, and the accuracy of model classification prediction is improved through an SMOTE improvement method; based on an SHAP method and partial dependency analysis, the relative importance degree of different input characteristic results and the positive and negative correlation influence and influence trend on different products in the DMO hydrogenation process are discussed; a classification result obtained by inputting characteristic data is returned to an APR2-ML regression device, the yields of different target products are optimized by adopting an APR2-ML model in combination with a genetic algorithm, and the characteristic parameters and reaction conditions of the catalyst are improved, so that the optimal catalyst can be screened out; by applying the model, novel catalysts with the best performance for preparing MG, EG and ET through DMO hydrogenation can be searched from a global optimization space, and catalyst multi-target optimization of a multi-target product process is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and in particular to a multi-objective product process control method and system based on machine learning. Background Art

[0002] The development of catalysts has always relied on a process of trial and error, which requires a lot of experimental time and resource investment. However, with the help of machine learning technology, the screening and optimization of catalysts have become more efficient. Machine learning can process complex, multi-dimensional data, and quickly identify potential candidates by predicting and simulating the performance of a large number of different catalyst combinations. This method is rapidly emerging in fields such as materials science and catalysis research. Traditional machine learning optimization and screening of catalysts generally goes through five steps: data collection and preprocessing, feature selection and extraction, model selection and training, model verification and testing, and catalyst performance prediction and screening.

[0003] Zahrt et al. (2019) proposed a computer-driven machine learning workflow for predicting highly selective catalysts, and believed that machine learning methods will provide chemists with a completely new paradigm for selecting and optimizing catalysts from experience-guided to mathematical-guided.

[0004] Maley et al. (2020) calculated 14 descriptors of the catalyst molecular structure for 105 Cr(P,N) catalysts, used density functional theory methods to calculate the transition state energy difference of catalytic ethylene polymerization to produce 1-hexene and 1-octene, and established a random forest regression model to predict the efficiency and selectivity of ethylene polymerization to produce specific products, and guide the design and optimization of catalysts for this reaction.

[0005] Chandanaetal. (2023) et al. developed an ML framework for modeling and predicting the catalyst activity of direct conversion of CO2 to light olefins C2–C4 based on structural composition and operating parameters. It includes 16 input parameters and catalyst activity as output parameters. Eight ML models were developed, including linear models, tree models, kernel-based models, and artificial neural network (ANN) models, and these models were used to predict catalyst activity.

[0006] However, most machine learning (ML) models used for catalyst development are designed for a single product and cannot adapt to the design of multi-target catalysts for processes such as the selective hydrogenation of dimethyl oxalate (DMO). In addition, traditional ML models mostly perform regression predictions based on the yield of the target product, making it difficult to predict and classify multi-target products based on catalyst characteristics and reaction condition parameters. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides a multi-objective product process control method and system based on machine learning, which solves the problem that multi-objective optimization of catalysts in multi-objective product processes is difficult to achieve.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A multi-objective product process control method based on machine learning, the method comprising the following steps:

[0010] S1. Collect data on catalyst properties and reaction conditions for multi-target product processes, study the distribution of the data, evaluate the data quality through data preprocessing and correlation analysis between features, and then input it into APR 2 -In ML models;

[0011] S2. Evaluate APR by micro-average, macro-average and AUC value under ROC curve 2 -ML classifier model and automatically screen the desired target products, and then automatically return the adaptive product data to APR through automated feedback 2 -In the ML model regressor, use the cross-validated determination coefficient and the root mean square error to calculate the APR 2 -Evaluate and improve the hyperparameters of ML regressors;

[0012] S3. Use SHAP-based feature importance analysis and partial dependence plots to explore the APR 2 -The influence degree and trend of input on output characteristics under ML, and analyze which characteristic parameters play an important role in automatic adjustment;

[0013] S4. Collect the multi-target product process catalyst product data selected through automatic classification and automatically feed it back to APR 2 -In the ML regressor, the optimization algorithm is combined to perform multi-parameter coordinated regulation and optimization of data such as catalyst properties and reaction conditions, and parameter data is automatically generated to maximize the yield of the final screened product;

[0014] S5. APR 2 -ML models are used in the multi-objective optimization screening and prediction of new catalysts for multi-objective product processes, automatically searching for potential candidate catalysts with better catalytic performance from the global optimization space.

[0015] Preferably, the multi-target product process is set as a DMO hydrogenation process.

[0016] Preferably, the APR 2 -The classifier of the ML model uses the RF model or the XGB model.

[0017] Preferably, the test set accuracy of the RF model is 0.85, and the recall rates on MG, EG, and ET are 0.80, 0.86, and 0.86;

[0018] The test set accuracy of the XGB model is 0.84, and the recall rates on MG, EG, and ET are 0.80, 0.85, and 0.86, respectively.

[0019] Preferably, the APR 2 -The regressor of the ML model adopts the XGB model.

[0020] Preferably, the determination coefficient of the XGB model on the test set is 0.86, and it has the best predictive ability for the yield of the product MG.

[0021] Preferably, through the TOPSIS decision method, it is concluded that the optimal catalyst for preparing MG by hydrogenation of DMO is Ag-Ni / SBA-15 loaded with 7.97wt% Ag; the optimal catalyst for preparing EG by hydrogenation of DMO is Cu-Au / SBA-15 loaded with 5.32wt% Cu; and the optimal catalyst for preparing ET by hydrogenation of DMO is Cu-HZ38 / SiO2 loaded with 12.55wt% Cu.

[0022] Preferably, the S5 comprises the following specific steps:

[0023] APR 2 -ML model combined with NSGA-II algorithm to find and predict the best new catalyst for DMO hydrogenation to MG, EG, ET in the global optimization space of catalyst properties and reaction conditions to maximize yield and minimize active ingredient loading;

[0024] The NSGA-II algorithm with a population size of 1000 and a maximum number of evolutions of 100 was used to search the feature space and serve as the APR 2 - Input of the ML model to calculate the function value of the target and set the mutation probability and crossover probability to 0.2 and 0.8;

[0025] The optimization model was run 100 times and the results of the 100 runs were compared to ensure the global optimal solution and to avoid falling into the trap of local minima during the spatial search process of the NSGA-II algorithm.

[0026] Preferably, by combining with the NSGA-II algorithm, APR 2The -ML model predicted six new catalysts for the preparation of MG, EG, and ET by hydrogenation of DMO, of which two Ag-based catalysts were used for the preparation of MG and four Cu-based catalysts were used for the preparation of EG and ET, namely Ag-B2O3 / HAP, Ag-CuO / KCC-1, Cu-Au / SNT, Cu-Al2O3 / SiO2, Cu-B / SNT, and Cu-Ni-foam / RGO.

[0027] A multi-objective product process control system based on machine learning, the system comprising: a data collection and evaluation module, a model selection and tuning module, an interpretability analysis module, a catalyst optimization module and a model application module;

[0028] The data collection and evaluation module is used to collect data related to catalyst properties and reaction conditions of multi-target product processes, and study the distribution of data. The data quality is evaluated through data preprocessing and correlation analysis between features and then input into APR 2 -In ML models;

[0029] The model selection and tuning module is used to evaluate the APR through micro-average and macro-average as well as the AUC value indicator under the ROC curve. 2 -ML classifier model and automatically screen the desired target products, and then automatically return the adaptive product data to APR through automated feedback 2 -In the ML model regressor, use the cross-validated determination coefficient and the root mean square error to calculate the APR 2 -Evaluate and improve the hyperparameters of ML regressors;

[0030] The interpretability analysis module is used to explore the features in APR using feature importance analysis based on SHAP values ​​and partial dependency plots. 2 -The influence degree and trend of input on output characteristics under ML, and analyze which characteristic parameters play an important role in automatic adjustment;

[0031] The catalyst optimization module is used to collect the multi-target product process catalyst product data selected through automatic classification and automatically feed it back to APR 2 -In the ML regressor, the optimization algorithm is combined to perform multi-parameter coordinated regulation and optimization of data such as catalyst properties and reaction conditions, and parameter data is automatically generated to maximize the yield of the final screened product;

[0032] The model application module is used to apply APR 2 -ML models are used in the multi-objective optimization screening and prediction of new catalysts for multi-objective product processes, automatically searching for potential candidate catalysts with better catalytic performance from the global optimization space.

[0033] The present invention provides a multi-objective product process control method and system based on machine learning. Compared with the prior art, it has the following beneficial effects:

[0034] In the present invention, the original data set is classified into three categories: methyl glycolate MG, ethylene glycol EG, and ethanol ET according to the difference in the size of the yield results. The accuracy of the model classification prediction is improved by the SMOTE improvement method, so that the machine learning model is more accurate in the result classification by inputting feature data without conducting experiments; based on the SHAP method and partial dependency analysis, the relative importance of different input feature results, as well as the positive and negative correlation effects and influence trends on different products of the DMO hydrogenation process are discussed; the classification results obtained by inputting feature data are returned to the APR 2 -In ML regressor, APR is used 2 -ML model combined with genetic algorithm can optimize the yield of different target products, improve catalyst characteristic parameters and reaction conditions, and screen out the optimal catalyst; the model can be used to search for the best new catalyst for DMO hydrogenation to MG, EG, and ET from the global optimization space, realizing multi-objective optimization of catalysts for multi-objective product processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 is a flow chart of the control method in an embodiment of the present invention;

[0037] Figure 2 The confusion matrices of the four models SVM (a), DNN (b), RF (c) and XGB (d) in Example 1 of the present invention are compared;

[0038] Figure 3 APR represented by MG for SVM, DNN, RF and XGB algorithms in Example 2 of the present invention 2 - Comparison of the predictive performance of the regressors of the ML models;

[0039] Figure 4 The Pareto set result of multi-objective optimization of the DMO hydrogenation catalyst in Example 3 of the present invention;

[0040] Figure 5 APR in Example 4 of the present invention 2-Comparison of six new optimal catalyst parameters predicted by ML model combined with genetic algorithm. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of 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.

[0042] The embodiments of the present application solve the problem that multi-objective optimization of catalysts in a multi-objective product process is difficult to achieve by providing a multi-objective product process control method and system based on machine learning.

[0043] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0044] Example:

[0045] In this embodiment, the multi-target product process takes the DMO hydrogenation process as an example.

[0046] like Figure 1 As shown, the present invention provides a multi-objective product process control method based on machine learning, and the method comprises the following steps:

[0047] S1. Data Collection and Evaluation

[0048] Collect relevant data on catalyst properties (CP) and reaction conditions (RC) of DMO hydrogenation process from a large number of experiments and literature, and study the distribution of data. After evaluating the data quality through data preprocessing and correlation analysis between features, the data is input into APR. 2 -In ML models;

[0049] S2. Model selection and tuning

[0050] APR is evaluated by micro-average, macro-average and AUC value under ROC curve 2 -ML classifier model and automatically screen the desired target products, and then automatically return the adaptive product data to APR through automated feedback 2 -In the ML model regressor, the cross-validated determination coefficient (R 2 ), the root mean square error (RMSE) of the APR 2 -Evaluate and improve the hyperparameters of ML regressors to improve APR 2 -ML model’s ability to adaptively control different target products;

[0051] S3. Interpretability Analysis

[0052] Explore the APR using feature importance analysis based on SHAP values ​​and partial dependence plots 2 -The influence degree and trend of input on output characteristics under ML, and analyze which characteristic parameters play an important role in automatic adjustment, so as to better adapt to product needs;

[0053] S4. Catalyst Optimization

[0054] Collect the DMO hydrogenation catalyst product data that has been automatically classified and screened, and automatically feed it back to APR 2 -In the ML regressor, the optimization algorithm is combined to perform multi-parameter coordinated regulation and optimization of CP and RC data, and the parameter data is automatically generated to maximize the yield of the DMO hydrogenation product substances finally screened out;

[0055] S5, APR 2 -Application of ML models

[0056] APR 2 -ML model is used in the multi-objective optimization screening and prediction of new DMO hydrogenation catalysts, automatically searching for potential candidate DMO hydrogenation catalysts with better catalytic performance from the global optimization space.

[0057] The present invention provides a multi-objective product process control system based on machine learning, the system comprising: a data collection and evaluation module, a model selection and tuning module, an interpretability analysis module, a catalyst optimization module and a model application module;

[0058] The data collection and evaluation module is used to collect data related to catalyst properties and reaction conditions of multi-target product processes, and study the distribution of data. The data quality is evaluated through data preprocessing and correlation analysis between features and then input into APR 2 -In ML models;

[0059] The model selection and tuning module is used to evaluate the APR through micro-average and macro-average as well as the AUC value indicator under the ROC curve. 2 -ML classifier model and automatically screen the desired target products, and then automatically return the adaptive product data to APR through automated feedback 2 -In the ML model regressor, use the cross-validated determination coefficient and the root mean square error to calculate the APR 2 -Evaluate and improve the hyperparameters of ML regressors to improve APR 2 -ML model’s ability to adaptively control different target products;

[0060] The interpretability analysis module is used to explore the features in APR using feature importance analysis based on SHAP values ​​and partial dependency plots. 2 -The influence degree and trend of input on output characteristics under ML, and analyze which characteristic parameters play an important role in automatic adjustment;

[0061] The catalyst optimization module is used to collect the multi-target product process catalyst product data selected through automatic classification and automatically feed it back to APR 2 -In the ML regressor, the optimization algorithm is combined to perform multi-parameter coordinated regulation and optimization of data such as catalyst properties and reaction conditions, and parameter data is automatically generated to maximize the yield of the final screened product;

[0062] The model application module is used to apply APR 2 -ML models are used in the multi-objective optimization screening and prediction of new catalysts for multi-objective product processes, automatically searching for potential candidate catalysts with better catalytic performance from the global optimization space.

[0063] Embodiment 1:

[0064] The test set Acy of RF and XGB models under the structures of n_estimators=71,max_depth=20,min_samples_split=2,min_samples_leaf=1 and n_estimators=86,max_depth=15,learning_rate=0.25,min_child_weight=8 are 0.85 and 0.84 respectively. The mean Acy of the two models under 5-fold cross validation is also as high as 0.84 and 0.83.

[0065] The classification performance of SVM and DNN models on this type of dataset is poor, with test set Acy of 0.8 and 0.81 respectively.

[0066] Therefore, the overall accuracy (Acy) index of the model preliminarily determined that RF and XGB are more suitable for APR than SVR and DNN ML models. 2 -Classifier of the ML model.

[0067] like Figure 2 As shown in the figure, the confusion matrix can more clearly show the accuracy of the four ML models under three different classification indicators: MG, EG and ET. The dark color value in the middle represents the recall rate (R). The higher the value, the more accurate the model classification.

[0068] observe Figure 2(ad) It can be found that the R index values ​​under the confusion matrix of the four ML models are significantly different. The R values ​​of RF and XGB under each index are higher than those of SVR and DNN models. Their R values ​​on MG, EG, and ET are 0.80, 0.86, 0.86 and 0.80, 0.85, 0.86, respectively. The highest R values ​​of SVR and DNN models on the three target predictions are 0.82 and 0.81, both of which are achieved in the classification task of predicting ET targets. It can be seen that RF and XGB models have significant advantages in capturing samples of different categories in this type of dataset.

[0069] Embodiment 2:

[0070] like Figure 3 As shown, the prediction performance of four models, SVM, DNN, RF, and XGB, is compared under DMO conversion rate. Figure 3 (a) shows the comparison results between the model prediction values ​​and the true values. It can be found that all models can fit the actual values ​​well in the prediction of the results, and the errors in the test set are mostly stable within 10%, indicating that the four models have a certain accuracy in predicting the results.

[0071] But according to Figure 3 (c) shows the results, XGB has the highest R in the final prediction 2 , the result on the test set is 0.86, so the yield Y of the product MG is MG The best prediction ability is in terms of 2 The R value of the test set is 0.84, while the performance of DNN and SVR on this type of data is not significantly better than that of the two integrated models. 2 They are 0.80 and 0.83 respectively.

[0072] pass Figure 3 The error distribution density shown in (d) also shows that the XGB model is more densely distributed near the error of 0, and its distribution frequency is the highest, close to 0.6, so XGB has the best prediction ability under this data set. Therefore, based on the above analysis, XGB is finally determined to be APR 2 -Optimal regressor for ML models.

[0073] Embodiment 3:

[0074] like Figure 4 As shown, we use the TOPSIS (Order Preference Technique of Similarity to an Ideal Solution) decision method to select the optimal value on the boundary. In Ag-based catalysts, such as Figure 4As shown in (ac), the TOPSIS method believes that Ag-Ni / SBA-15 (82.60%, 7.97wt%), Ag-Ti / KCC-1 (71.73%, 9.95wt%), and Ag-B2O3 / SiO2 (76.94%, 12.62wt%) are the best catalysts for DMO hydrogenation to MG, EG, and ET, respectively. It can be found that Ag-based catalysts have significant advantages in DMO hydrogenation to MG. The MG yield after optimization is 82.60%, which is higher than 71.73% and 76.94% of EG and ET, and only requires a lower active component loading of 7.97wt%. For Cu-based catalysts, Figure 3 As shown in (df), the TOPSIS method believes that Cu-Al2O3 / ZrO2 (77.59%, 13.15wt%), Cu-Au / SBA-15 (83.20%, 5.32wt%), and Cu-HZ38 / SiO2 (82.60%, 12.55wt%) are the best catalysts for DMO hydrogenation to prepare MG, EG, and ET, respectively. The Cu-based catalyst shows good results in the preparation of EG, and its performance in yield (83.20%) and active component loading (5.32wt%) are better than MG and ET. Secondly, the performance of the Cu-based catalyst on ET is also relatively excellent, and TOPSIS believes that (82.60%, 12.55wt%) is also a relatively advantageous choice. Overall, the Ag catalyst is more suitable for the process of DMO hydrogenation to prepare MG, while the Cu-based catalyst has more advantages in the process of preparing EG and ET.

[0075] Embodiment 4:

[0076] like Figure 5 As shown, using APR 2 -ML model is combined with NSGA-II algorithm to find and predict the best new catalyst for DMO hydrogenation to MG, EG, and ET in the global optimization space of catalyst properties and reaction conditions to maximize yield and minimize active component loading.

[0077] The NSGA-II algorithm with a population size of 1000 and a maximum number of evolutions of 100 was used to search the feature space and serve as the APR 2 - Input to the ML model to calculate the function value of the target and set the mutation and crossover probabilities to 0.2 and 0.8.

[0078] The optimization model was run 100 times and the results of the 100 runs were compared to ensure the global optimal solution and to avoid falling into the trap of local minima during the spatial search process of the NSGA-II algorithm.

[0079] Finally, by combining with the NSGA-II algorithm, APR 2 -ML model successfully predicted 6 new catalysts for DMO hydrogenation to prepare MG, EG, and ET, of which two Ag-based catalysts were used for the preparation of MG, and four Cu-based catalysts were used for the preparation of EG and ET. They are Ag-B2O3 / HAP, Ag-CuO / KCC-1, Cu-Au / SNT, Cu-Al2O3 / SiO2, Cu-B / SNT, Cu-Ni-foam / RGO, Figure 5 (a)-(b) Their catalytic performances were compared.

[0080] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0081] In the embodiment of the present invention, the original data set is classified into three categories: methyl glycolate MG, ethylene glycol EG, and ethanol ET according to the difference in the size of the yield results. The accuracy of the model classification prediction is improved by the SMOTE improvement method, so that the machine learning model is more accurate in the result classification by inputting feature data without conducting experiments; based on the SHAP method and partial dependency analysis, the relative importance of different input feature results, as well as the positive and negative correlation effects and influence trends on different products of the DMO hydrogenation process are discussed; the classification results obtained by inputting feature data are returned to the APR 2 -In ML regressor, APR is used 2 -ML model combined with genetic algorithm can optimize the yield of different target products, improve catalyst characteristic parameters and reaction conditions, and screen out the optimal catalyst; the model can be used to search for the best new catalyst for DMO hydrogenation to MG, EG, and ET from the global optimization space, realizing multi-objective optimization of catalysts for multi-objective product processes.

[0082] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-objective product process control method based on machine learning, characterized in that: The method comprises the following steps: S1. Collect data on catalyst properties and reaction conditions for multi-target product processes, study the distribution of the data, evaluate the data quality through data preprocessing and correlation analysis between features, and then input it into APR 2 -In ML models; S2. Evaluate APR by micro-average, macro-average and AUC value under ROC curve 2 -ML classifier model automatically screens the desired target products and then automatically returns the adaptive product data to APR through automated feedback 2 -In the ML model regressor, use the cross-validated determination coefficient and the root mean square error to calculate the APR 2 -Evaluate and improve the hyperparameters of ML regressors; S3. Use SHAP value-based feature importance analysis and partial dependence plots to explore the APR 2 -The influence degree and trend of input on output characteristics under ML, and analyze which characteristic parameters play an important role in automatic adjustment; S4. Collect the multi-target product process catalyst product data selected through automatic classification and automatically feed it back to APR 2 -In the ML regressor, the optimization algorithm is combined to perform multi-parameter coordinated regulation and optimization of data such as catalyst properties and reaction conditions, and parameter data is automatically generated to maximize the yield of the final screened product; S5. APR 2 -ML models are used in the multi-objective optimization screening and prediction of new catalysts for multi-objective product processes, automatically searching for potential candidate catalysts with better catalytic performance from the global optimization space.

2. The multi-objective product process control method based on machine learning according to claim 1, characterized in that: The multi-target product process is set as a DMO hydrogenation process.

3. The multi-objective product process control method based on machine learning according to claim 2, characterized in that: The APR 2 -The classifier of the ML model uses the RF model or the XGB model.

4. The multi-objective product process control method based on machine learning according to claim 3, characterized in that: The test set accuracy of the RF model is 0.85, and the recall rates on MG, EG, and ET are 0.80, 0.86, and 0.86; The test set accuracy of the XGB model is 0.84, and the recall rates on MG, EG, and ET are 0.80, 0.85, and 0.86, respectively.

5. The multi-objective product process control method based on machine learning according to claim 2, characterized in that: The APR 2 -The regressor of the ML model adopts the XGB model.

6. The multi-objective product process control method based on machine learning according to claim 5, characterized in that: The determination coefficient of the XGB model on the test set is 0.86, and it has the best predictive ability for the yield of the product MG.

7. The multi-objective product process control method based on machine learning according to claim 2, characterized in that: Through the TOPSIS decision-making method, it was concluded that the optimal catalyst for DMO hydrogenation to prepare MG is Ag-Ni / SBA-15 loaded with 7.97wt% Ag; the optimal catalyst for DMO hydrogenation to prepare EG is Cu-Au / SBA-15 loaded with 5.32wt% Cu; and the optimal catalyst for DMO hydrogenation to prepare ET is Cu-HZ38 / SiO2 loaded with 12.55wt% Cu.

8. The multi-objective product process control method based on machine learning according to claim 2, characterized in that: The S5 comprises the following specific steps: APR 2 -ML model combined with NSGA-II algorithm to find and predict the best new catalyst for DMO hydrogenation to MG, EG, ET in the global optimization space of catalyst properties and reaction conditions to maximize yield and minimize active ingredient loading; The NSGA-II algorithm with a population size of 1000 and a maximum number of evolutions of 100 was used to search the feature space and serve as the APR 2 - Input of the ML model to calculate the function value of the target and set the mutation probability and crossover probability to 0.2 and 0.8; The optimization model was run 100 times and the results of the 100 runs were compared to ensure the global optimal solution and to avoid falling into the trap of local minima during the spatial search process of the NSGA-II algorithm.

9. The multi-objective product process control method based on machine learning according to claim 8, characterized in that: According to the combination with NSGA-II algorithm, APR 2 The -ML model predicted six new catalysts for the preparation of MG, EG, and ET by hydrogenation of DMO, of which two Ag-based catalysts were used for the preparation of MG and four Cu-based catalysts were used for the preparation of EG and ET, namely Ag-B2O3 / HAP, Ag-CuO / KCC-1, Cu-Au / SNT, Cu-Al2O3 / SiO2, Cu-B / SNT, and Cu-Ni-foam / RGO.

10. A multi-objective product process control system based on machine learning, characterized in that: The system includes: a data collection and evaluation module, a model selection and tuning module, an interpretability analysis module, a catalyst optimization module and a model application module; The data collection and evaluation module is used to collect data related to catalyst properties and reaction conditions of multi-target product processes, and study the distribution of data. The data quality is evaluated through data preprocessing and correlation analysis between features and then input into APR 2 -In ML models; The model selection and tuning module is used to evaluate the APR through micro-average and macro-average as well as the AUC value indicator under the ROC curve. 2 -ML classifier model automatically screens the desired target products and then automatically returns the adaptive product data to APR through automated feedback 2 -In the ML model regressor, use the cross-validated determination coefficient and the root mean square error to calculate the APR 2 -Evaluate and improve the hyperparameters of ML regressors; The interpretability analysis module is used to explore the features in APR using feature importance analysis based on SHAP values ​​and partial dependency plots. 2 -The influence degree and trend of input on output characteristics under ML, and analyze which characteristic parameters play an important role in automatic adjustment; The catalyst optimization module is used to collect the multi-target product process catalyst product data selected through automatic classification and automatically feed it back to APR 2 -In the ML regressor, the optimization algorithm is combined to perform multi-parameter coordinated regulation and optimization of data such as catalyst properties and reaction conditions, and parameter data is automatically generated to maximize the yield of the final screened product; The model application module is used to apply APR 2 -ML models are used in the multi-objective optimization screening and prediction of new catalysts for multi-objective product processes, automatically searching for potential candidate catalysts with better catalytic performance from the global optimization space.

Images