A platinum group nanomaterial catalyst formulation optimization method and system

Abstract

The present application proposes a method and system for optimizing the formulation of platinum group nanomaterial catalysts, which belongs to the field of computer-aided material design. Among them, the method includes: constructing a multi-objective optimization space for formulation performance, and constructing a fitness function based on the weight distribution of performance indicators; randomly configuring a number of catalyst formulations, combining the preparation condition parameter index sample set, and obtaining a high-frequency catalyst structure; processing the catalyst structure through a pre-trained catalyst performance mapping model, obtaining a performance indicator prediction value and inputting the fitness function to obtain the fitness; performing optimization based on the fitness, determining the target catalyst formulation and sending it to the formulation control end. The present invention solves the technical problem that the optimization of platinum group nanomaterial catalyst formulations in the prior art does not take into account the differentiated performance requirements of different application scenarios, resulting in low formulation optimization efficiency and poor applicability, and achieves the technical effect of improving formulation optimization efficiency and applicability, and realizing differentiated formulation optimization for different application scenarios.

Description

Technical Field

[0001] The present invention relates to the field of computer-aided material design, and in particular to a method and system for optimizing a platinum group nanomaterial catalyst formulation. Background Art

[0002] Platinum-group nanomaterial catalysts, due to their excellent catalytic activity, selectivity, and stability, hold broad application prospects in fields such as petrochemicals, environmental remediation, and new energy. With the increasing demand for industrial applications, the performance requirements for platinum-group nanomaterial catalysts are becoming increasingly stringent. Different application scenarios place varying emphasis on performance indicators such as catalytic activity and cyclic stability.

[0003] Traditional optimization of platinum group nanomaterial catalyst formulas mainly relies on trial and error and experience accumulation. This method is not only time-consuming and labor-intensive, but also costly, and it is difficult to achieve coordinated optimization of multiple performance indicators. Existing formula optimization methods usually adopt a single optimization target or a fixed weight distribution method, which fails to fully consider the differentiated requirements of catalyst performance in different application scenarios. For example, in automobile exhaust treatment applications, more attention is paid to the low-temperature activity and sulfur poisoning resistance of the catalyst, while in fuel cell applications, more attention is paid to the mass activity and long-term stability of the catalyst. In addition, traditional optimization methods lack effective performance prediction methods and often require a lot of experimental verification to determine the optimal formula, which not only prolongs the R&D cycle, but also limits the efficiency of formula optimization. Therefore, the optimization of platinum group nanomaterial catalyst formulas in the existing technology does not take into account the differentiated performance requirements of different application scenarios, resulting in low formula optimization efficiency and poor applicability. Summary of the Invention

[0004] The present invention aims to solve the technical problem that the formulation optimization of platinum group nanomaterial catalysts in the prior art does not take into account the differentiated performance requirements of different application scenarios, resulting in low formulation optimization efficiency and poor applicability. A method and system for optimizing the formulation of platinum group nanomaterial catalysts are provided to solve the problem.

[0005] The technical solution of the present invention to solve the above technical problems is as follows:

[0006] In a first aspect, the present invention provides a method for optimizing a platinum group nanomaterial catalyst formulation, comprising: constructing a multi-objective optimization space for formulation performance according to predefined performance indicator boundary values, weighting preset performance indicators based on a predefined performance indicator weight distribution, and constructing a fitness function belonging to the multi-objective optimization space for formulation performance, wherein the preset performance indicators are positive indicators; randomly configuring a number of platinum group nanomaterial catalyst formulations, combining preparation condition parameters, indexing a catalyst preparation sample set, and obtaining a number of high-frequency catalyst structures; processing the several high-frequency catalyst structures separately through a pre-trained catalyst performance mapping model, obtaining a number of performance indicator prediction values, inputting the fitness function, and obtaining a number of fitnesses; performing optimization based on the several fitnesses and the several platinum group nanomaterial catalyst formulations, determining a target catalyst formulation, and sending it to a catalyst formulation management and control end.

[0007] In a second aspect, the present invention provides a platinum group nanomaterial catalyst formula optimization system, comprising: an optimization parameter construction module, used to construct a multi-objective optimization space for formula performance according to predefined performance indicator boundary values, weight the preset performance indicators based on the predefined performance indicator weight distribution, and construct a fitness function belonging to the multi-objective optimization space for formula performance, wherein the preset performance indicators are positive indicators; a formula generation module, used to randomly configure a number of platinum group nanomaterial catalyst formulas, combine preparation condition parameters, index the catalyst preparation sample set, and obtain a number of high-frequency catalyst structures; a fitness calculation module, used to process the several high-frequency catalyst structures respectively through a pre-trained catalyst performance mapping model, obtain a number of performance indicator prediction values, input the fitness function, and obtain a number of fitnesses; an optimization output module, used to perform optimization based on the several fitnesses and the several platinum group nanomaterial catalyst formulas, determine the target catalyst formula, and send it to the catalyst formula management and control end.

[0008] The beneficial effects of the present invention are:

[0009] Based on the predefined performance indicator boundary values, a multi-objective optimization space for formulation performance was constructed. The preset performance indicators were weighted based on the predefined performance indicator weight distribution, and a fitness function belonging to the multi-objective optimization space for formulation performance was constructed. The preset performance indicators were used as positive indicators. This established a performance evaluation system for different application scenarios, providing support for subsequent formulation optimization. Several platinum-group nanomaterial catalyst formulations were randomly configured. Combined with preparation condition parameters, the catalyst preparation sample set was indexed to obtain several high-frequency catalyst structures, and a representative candidate formulation space was constructed, providing a diverse structural foundation for performance prediction. Using a pre-trained catalyst performance mapping model, several high-frequency catalyst structures were processed separately to obtain several performance indicator prediction values. These values ​​were then input into the fitness function to obtain several fitness values. This achieved a rapid mapping from catalyst structure to performance indicator, avoiding extensive experimental verification. The fitness function also converted multiple performance indicators into comparable optimization targets. Optimization was performed based on several fitness values ​​and several platinum-group nanomaterial catalyst formulations to determine the target catalyst formulation, which was then sent to the catalyst formulation control terminal. This implemented intelligent optimization based on fitness, determined the optimal formulation that met the requirements of a specific application scenario, and achieved automated output of the formulation.

[0010] By constructing a multi-objective optimization space and fitness function, the optimization objectives can be flexibly adjusted according to the performance requirements of different application scenarios. The use of pre-trained performance mapping models replaces a large number of experimental verifications, significantly shortening the formulation optimization cycle. The intelligent optimization algorithm achieves the coordinated optimization of multiple performance indicators, avoiding the performance imbalance caused by single-indicator optimization in traditional methods. Therefore, through the above technical solutions, this application improves the efficiency and applicability of platinum group nanomaterial catalyst formulation optimization and realizes differentiated formulation optimization for different application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic flow chart of a method for optimizing the formulation of a platinum group nanomaterial catalyst provided by the present invention;

[0012] Figure 2 This is a schematic structural diagram of a platinum group nanomaterial catalyst formulation optimization system provided by the present invention.

[0013] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0014] Optimization parameter construction module 11, recipe generation module 12, fitness calculation module 13, optimization output module 14. DETAILED DESCRIPTION

[0015] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0016] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0017] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0018] Example 1, as Figure 1 As shown, an embodiment of the present invention provides a method for optimizing a platinum group nanomaterial catalyst formulation, comprising:

[0019] S1. According to the predefined performance indicator boundary values, a multi-objective optimization space for recipe performance is constructed. Based on the predefined performance indicator weight distribution, the preset performance indicators are weighted to construct a fitness function belonging to the multi-objective optimization space for recipe performance, where the preset performance indicators are positive indicators.

[0020] Specifically, a multi-objective optimization space for formulation performance is first constructed based on predefined performance indicator boundary values. These performance indicator boundary values ​​are determined based on domain expertise and historical experimental data, defining the acceptable range for each performance indicator and thus limiting the effective region for optimization search. These performance indicator boundary values ​​correspond to multiple preset performance indicators, including but not limited to catalytic activity, cycle count, and stability. The setting of performance indicator boundary values ​​ensures that the optimization process focuses on formulation regions with practical application value, avoiding unnecessary searches in invalid solution spaces.

[0021] The multi-objective optimization space for formulation performance is a mathematical construct that forms a multidimensional solution space containing all possible catalyst formulations and their corresponding performance index values. Specifically, this space consists of multiple dimensions, each representing a preset performance index (such as catalytic activity, number of cycles, stability, etc.), and performance index boundary values ​​are set in each dimension to limit the effective search area. In this space, each feasible catalyst formulation is represented as a multi-dimensional coordinate point, and the core optimization goal is to find the optimal balance between these conflicting performance indicators. The construction of the multi-objective optimization space for formulation performance enables the effective identification of the optimal solution, providing support for the search for the target catalyst formulation.

[0022] Secondly, based on the predefined performance indicator weight distribution, the preset performance indicators are weighted to construct a fitness function belonging to the multi-objective optimization space of the formulation performance. Among them, the performance indicator weight distribution reflects the relative importance of different performance indicators in a specific application scenario. Through this weighting mechanism, the optimization direction can be flexibly adjusted according to the specific needs of different application scenarios, thereby achieving targeted catalyst formulation optimization. The fitness function is a comprehensive evaluation indicator that integrates multiple preset performance indicators into a single evaluation standard through a mathematical model, which is used to quantify the comprehensive performance of different catalyst formulations. Among them, the preset performance indicators are all positive indicators, that is, the larger the indicator value, the better the catalyst performance. This positive treatment simplifies the construction process of the fitness function and makes the optimization goal clear as a fitness value maximization problem.

[0023] Through the above technical solution, a multi-dimensional and comprehensive evaluation of the performance of platinum group nanomaterial catalyst formulations was achieved, laying a mathematical foundation for subsequent formulation optimization.

[0024] S2. Randomly configure several platinum group nanomaterial catalyst formulas, combine the preparation condition parameters, index the catalyst preparation sample set, and obtain several high-frequency catalyst structures.

[0025] Specifically, they first randomly assigned several platinum-group nanomaterial catalyst formulations within a multi-objective optimization space. This random assignment strategy ensures extensive exploration within the multi-objective optimization space, avoiding excessive focus on specific regions and thus increasing the likelihood of discovering new high-performance catalysts.

[0026] Next, the catalyst preparation sample set is indexed based on the preparation parameters. Specifically, the same platinum-group nanomaterial catalyst formulation can produce different microstructures under different preparation conditions, and thus exhibit different catalytic performance. Therefore, the dual influence of formulation composition and preparation conditions must be considered during the optimization process. By combining a randomly configured platinum-group nanomaterial catalyst formulation with specific preparation parameters, the corresponding catalyst preparation records are retrieved from the historical experimental database to obtain the corresponding catalyst preparation sample set.

[0027] Then, by performing structural similarity cluster analysis on the catalyst preparation records in the retrieved catalyst preparation sample set, we identified the catalyst structure types that may form under given formulations and preparation conditions, and obtained several high-frequency catalyst structures. These high-frequency structures represent catalyst morphologies with high reproducibility and stability in actual preparation processes, providing a reliable structural foundation for subsequent performance prediction and optimization.

[0028] By obtaining several high-frequency catalyst structures, an effective mapping from catalyst formula to actual structure is achieved, which solves the problem of the difference between theoretical design and actual preparation in catalyst research and development, and improves the practical application value and reliability of the optimization results.

[0029] S3. Using a pre-trained catalyst performance mapping model, the plurality of high-frequency catalyst structures are processed respectively to obtain a plurality of performance index prediction values, which are input into the fitness function to obtain a plurality of fitness values.

[0030] Specifically, the team first processed each of the several high-frequency catalyst structures using a pre-trained catalyst performance mapping model. This pre-trained model, a machine learning model trained on extensive historical experimental data, accurately predicts various catalyst performance indicators based on their structural characteristics. This model enables rapid performance evaluation of various high-frequency catalyst structures, avoiding tedious experimental verification and improving optimization efficiency.

[0031] For each high-frequency catalyst structure obtained by S2, its structural parameters are input into a pre-trained catalyst performance mapping model for forward calculation, resulting in corresponding performance index predictions, including catalytic activity, cycle number, and stability. These performance index predictions reflect the performance of each high-frequency catalyst structure across various performance dimensions, providing basic data for comprehensive evaluation.

[0032] The predicted performance indicators are then substituted into the fitness function constructed in S1 to calculate the comprehensive fitness value for each high-frequency catalyst structure. These comprehensive fitness values ​​comprehensively consider the weight distribution of each performance indicator and can fully reflect the comprehensive performance of the catalyst in a specific application scenario.

[0033] By obtaining the fitness of several high-frequency catalyst structures, rapid performance evaluation and comprehensive fitness calculation of the catalyst structure are achieved, providing a reliable quantitative basis for subsequent optimization decisions and effectively solving the problems of time-consuming and high-cost performance evaluation in traditional methods.

[0034] S4. Optimizing the plurality of fitnesses and the plurality of platinum group nanomaterial catalyst formulations is performed to determine a target catalyst formulation, and the target catalyst formulation is sent to a catalyst formulation management and control terminal.

[0035] Specifically, an optimization search is first performed based on several fitness values ​​and several platinum-group nanomaterial catalyst formulations. Specifically, the obtained fitness values ​​are used to iteratively optimize the multiple platinum-group nanomaterial catalyst formulations within a multi-objective optimization space for formulation performance. During this optimization process, platinum-group nanomaterial catalyst formulations with higher fitness values ​​exert a stronger attraction on other surrounding platinum-group nanomaterial catalyst formulations, guiding the search process to converge toward regions of high fitness while also retaining a certain degree of global exploration capability to avoid being trapped in local optimal solutions. Through multiple rounds of iterative calculations, the platinum-group nanomaterial catalyst formulations are gradually adjusted and optimized, continuously improving their fitness values ​​until the preset convergence conditions are reached.

[0036] When the optimization process reaches convergence, such as reaching a preset number of convergence iterations or satisfying other termination criteria, the platinum-group nanomaterial catalyst formulation with the highest fitness is selected as the target catalyst formulation. This target formulation represents the platinum-group nanomaterial catalyst formulation with the best overall performance indicators for a given application scenario.

[0037] The optimized target catalyst formula is then sent to the catalyst formula control terminal, which can be a production control system, laboratory management system, or other related system. This terminal is responsible for converting the optimization results (i.e., the target catalyst formula) into actual catalyst preparation instructions and process parameters.

[0038] By dynamically adjusting the weights of performance indicators according to the needs of specific application scenarios and optimizing catalyst formulations in a targeted manner, the efficiency and applicability of formulation optimization are significantly improved, providing support for customized high-performance catalyst formulations for different application scenarios.

[0039] Furthermore, according to the predefined performance indicator boundary values, a multi-objective optimization space for recipe performance is constructed, and based on the predefined performance indicator weight distribution, the preset performance indicators are weighted to construct a fitness function belonging to the multi-objective optimization space for recipe performance, which includes:

[0040] S11, receiving a target application scenario, performing a correlation analysis on preset performance indicators of the platinum group nanomaterial catalyst, and obtaining a weight distribution of the performance indicators;

[0041] S12. Counting a sample set of catalysts for the target application scenario, performing a mode analysis on the preset performance indicators of the platinum group nanomaterial catalyst, and obtaining the boundary values ​​of the performance indicators.

[0042] In a feasible implementation, the performance indicator weight distribution and the performance indicator boundary value are adaptively determined according to the target application scenario.

[0043] First, the target application scenario is received, and based on the received target application scenario, a correlation analysis is performed on the preset performance indicators of the platinum group nanomaterial catalyst. Among them, the target application scenario refers to the specific industrial environment or chemical reaction process in which the platinum group nanomaterial catalyst is actually applied, including but not limited to automobile exhaust purification, fuel cell electrode reaction, petroleum catalytic cracking, fine chemical synthesis and other different fields. Due to differences in its reaction conditions, reactant composition, target product and process requirements, each application scenario has different requirements, preferences and importance for performance indicators such as catalyst activity and stability, which is the basis for achieving differentiated optimization. The correlation analysis quantitatively evaluates the degree of influence of each performance indicator on the target application scenario by examining the correlation between each performance indicator and the effect indicator of the target application scenario. Through the correlation calculation, the performance indicator weight distribution is obtained. The performance indicator weight distribution reflects the relative importance of each preset performance indicator in a specific application scenario, providing a basis for the subsequent construction of the fitness function.

[0044] Next, a sample set of catalysts for the target application scenario was compiled. This sample set contained records of catalysts that were actually used in that target application scenario and produced qualified target products. Based on this valuable historical data, a mode analysis was performed on the various preset performance indicators of platinum-group nanomaterial catalysts to identify the concentrated distribution intervals and effective ranges of each preset performance indicator. This data-driven analysis process yielded performance indicator boundary values. These boundary values ​​defined the effective range of each performance indicator in the target application scenario, providing clear boundary constraints for constructing a multi-objective optimization space for formulation performance.

[0045] Through the above steps, the adaptive determination of the performance indicator weight distribution and performance indicator boundary values ​​is achieved, so that the subsequently constructed multi-objective optimization space and fitness function of the formula performance can accurately reflect the demand characteristics of specific application scenarios, laying the foundation for the targeted optimization of platinum group nanomaterial catalyst formulas.

[0046] Furthermore, the target application scenario is received, and a correlation analysis is performed on the preset performance indicators of the platinum group nanomaterial catalyst to obtain the weight distribution of the performance indicators, including:

[0047] S111. Extracting user-preconfigured evaluation trait attributes from the target application scenario;

[0048] S112. Based on the target application scenario, collect one-to-one correspondences of: evaluation trait attribute characteristic value sequence, first preset performance indicator characteristic value sequence, up to Nth preset performance indicator characteristic value sequence;

[0049] S113, using the evaluation trait attribute characteristic value sequence as a reference sequence and the first preset performance indicator characteristic value sequence to the Nth preset performance indicator characteristic value sequence as a comparison sequence, performing grey relational analysis to generate first preset performance indicator correlation degrees to Nth preset performance indicator correlation degrees;

[0050] S114 , traverse the first preset performance indicator correlation degree until the Nth preset performance indicator correlation degree, add and compare with the correlation degree, and obtain the performance indicator weight distribution.

[0051] In a feasible implementation, each application scenario has its own specific evaluation trait attributes, which are pre-configured by domain experts or users based on application requirements. First, by parsing the configuration information of the target application scenario, these evaluation trait attributes, such as reaction conversion rate, product selectivity, energy consumption index, etc., are extracted as benchmark indicators for evaluating catalyst performance. For example, in the automobile exhaust purification scenario, it may include NOx conversion rate, CO conversion rate, etc., and in the fuel cell scenario, it may include electrode reaction activity, potential stability, etc. These evaluation trait attributes represent the user's performance expectations for the catalyst in a specific application scenario and are the basis for correlation analysis.

[0052] Then, based on the target application scenario, multiple sets of corresponding characteristic value sequences are collected from the historical database. Specifically, for the same batch of historical experimental data, the measured values ​​of the evaluation properties are extracted to form the evaluation property characteristic value sequence. At the same time, the preset performance index values ​​of the catalysts in these experiments are extracted to form the first preset performance index characteristic value sequence through the Nth preset performance index characteristic value sequence. For example, for N preset performance indicators such as catalytic activity index, cycle number index, and stability index, N preset performance index characteristic value sequences are constructed respectively. Each sequence corresponds one-to-one with a data point in the evaluation property characteristic value sequence, reflecting the measurement results of different indicators under the same experimental conditions.

[0053] Subsequently, the characteristic value sequence of the evaluation trait attribute is set as the baseline sequence, and the characteristic value sequence of each preset performance indicator is set as the comparison sequence. The gray correlation degree between each comparison sequence and the baseline sequence is then calculated. Specifically, each sequence is first dimensionlessly processed to eliminate the influence of dimensional differences between different indicators and ensure the comparability of the analysis results. Second, the absolute value of the difference between the corresponding points in the baseline sequence and each comparison sequence is calculated to form a difference sequence. Then, based on the difference sequence, the correlation coefficient of each point is calculated. The correlation coefficient reflects the degree of similarity between each point in the comparison sequence and the corresponding point in the baseline sequence. After that, the arithmetic mean of the correlation coefficients of each point is calculated to obtain the correlation degree of the first preset performance indicator through the Nth preset performance indicator. These correlation values ​​quantitatively reflect the degree of influence of each preset performance indicator on the evaluation trait attribute. The higher the correlation degree, the stronger the correlation between the performance indicator and the application effect, the greater its contribution to the application effect, and thus should be given a higher weight in subsequent optimization.

[0054] Next, the correlations of the first preset performance indicator are traversed through the Nth preset performance indicator correlations. The sum of all correlations is calculated, and then each preset performance indicator correlation is divided by this sum to obtain the weight of each preset performance indicator. This normalization process ensures that the sum of all weights is 1 while maintaining the relative proportional relationship between the correlations of each indicator, thereby obtaining a performance indicator weight distribution. This weight distribution accurately reflects the relative importance of each preset performance indicator in a specific target application scenario and provides a basis for the subsequent construction of the fitness function.

[0055] Through the above steps, the weight distribution of each preset performance indicator can be determined according to the specific needs of the target application scenario, so that the subsequent optimization process can accurately reflect the differentiated requirements of different application scenarios for catalyst performance, thereby improving the targeted optimization of catalyst formulations.

[0056] Furthermore, a catalyst sample set of the target application scenario is statistically analyzed, and a mode analysis is performed on the preset performance indicators of the platinum group nanomaterial catalyst to obtain the boundary values ​​of the performance indicators, including:

[0057] S121. Collect a catalyst sample set with target product quality that meets the target application scenario, where the target product quality refers to the substance that needs to be produced by the catalyst;

[0058] S122. Extracting a first set of preset performance indicator recorded values ​​based on the catalyst sample set, performing box plot analysis to obtain a first set of preset performance indicator recorded values, and constructing a first preset performance indicator boundary value using the maximum and minimum values ​​of the first set of preset performance indicator recorded values;

[0059] S123, until the Nth preset performance indicator record value set is extracted based on the catalyst sample set, a box plot analysis is performed to obtain the preset performance indicator record values ​​in the Nth set, and the Nth preset performance indicator boundary value is constructed using the maximum and minimum values ​​of the preset performance indicator record values ​​in the Nth set.

[0060] In a preferred embodiment, in the process of performing mode analysis on the preset performance indicators of the platinum group nanomaterial catalyst to obtain the boundary values ​​of the performance indicators, the effective value range of each preset performance indicator is determined by statistical analysis based on qualified cases in the target application scenario.

[0061] First, a sample set of catalysts with qualified target product quality for the target application scenario is collected. Target product quality refers to the quality characteristics of the substance produced by the catalyst, such as purity, yield, and stability. Catalyst records that produce qualified target products in specific target application scenarios are screened from the historical database to form a catalyst sample set. This represents catalyst cases that have been proven effective in real-world applications and serves as a reliable data source for determining performance indicator boundaries.

[0062] Then, based on the catalyst sample set, the first preset performance indicator record value set is extracted, and a box plot analysis is performed. Specifically, for the first preset performance indicator (such as the catalytic activity index), all relevant record values ​​are extracted from the catalyst sample set to form a first preset performance indicator record value set. Subsequently, a box plot analysis is performed on the set to display the distribution characteristics of the first preset performance indicator record value, including the median, quartiles, and possible outliers in the first preset performance indicator record value. Through the box plot analysis, the preset performance indicator record values ​​in the first set are identified, that is, the main distribution range after excluding outliers. Then, the maximum and minimum values ​​of the preset performance indicator record values ​​in the first set are used to construct the first preset performance indicator boundary value, which defines the effective value range of the first preset performance indicator in the target application scenario.

[0063] Repeat the process of S122, processing the second through the Nth preset performance indicators in sequence. For each preset performance indicator, extract its set of recorded values, perform box plot analysis, obtain a centralized set of recorded values, and construct the boundary values ​​for that indicator using its maximum and minimum values. In this way, a complete set of boundary values ​​from the first through the Nth preset performance indicator is obtained, comprehensively defining the valid value space for each performance indicator in the target application scenario.

[0064] Through the above steps, adaptive determination of performance indicator boundary values ​​based on the target application scenario is achieved, allowing the optimization search space to be precisely focused on formulation ranges with practical application value, avoiding unnecessary searches in the invalid solution space, thereby improving the efficiency and success rate of catalyst formulation optimization. Compared with traditional empirical settings or global searches, this data-driven boundary value determination method is more targeted and reliable, providing a basis for precise optimization of catalyst formulations.

[0065] Furthermore, the mathematical model of the fitness function is as follows:

[0066] ,

[0067] ,

[0068] in, Characterize the fitness value of any platinum group nanomaterial catalyst formula, Characterize the predicted value of the performance index of platinum group nanomaterial catalyst formula, x represents any platinum group nanomaterial catalyst formula, Characterizes the minimum boundary value of the preset performance indicator of the i-th attribute, Characterizes the maximum boundary value of the preset performance indicator of the i-th attribute, Characterizes the preset performance indicator weight of the i-th attribute.

[0069] In a preferred embodiment, the mathematical model of the fitness function consists of two main parts:

[0070] ,

[0071] ,

[0072] in, Characterizing the fitness value of any platinum group nanomaterial catalyst formula is a quantitative evaluation of the comprehensive performance of the catalyst formula. represents any platinum group nanomaterial catalyst formula, where The predicted values ​​of various performance indicators that characterize the platinum-group nanomaterial catalyst formula are obtained through a pre-trained catalyst performance mapping model.

[0073] Among them, the fitness function A weighted normalization calculation method is used. For each preset performance indicator , first according to its minimum boundary value and the maximum boundary value Normalization is performed to convert the index values ​​of different dimensions and magnitudes into standardized values ​​within the interval [0,1]. Calculate the normalized value of each preset performance indicator and then multiply it by the corresponding weight , and finally sum up all weighted normalized values ​​to get the final fitness value. The minimum boundary value of the preset performance indicator representing the i-th attribute is determined by the corresponding preset performance indicator boundary value; The maximum boundary value of the preset performance indicator representing the i-th attribute is also determined by the corresponding preset performance indicator boundary value; The weight of the preset performance indicator representing the i-th attribute is determined by the performance indicator weight distribution. This boundary value and weight setting mechanism ensures that the fitness function can adaptively adjust the evaluation criteria according to the specific needs of different application scenarios.

[0074] Through the above-mentioned fitness function, the evaluation and quantitative comparison of platinum group nanomaterial catalyst formulas are achieved, providing a clear objective function for subsequent formula optimization, so that the optimization process can be effectively carried out in the direction of improving the overall fitness value, thereby improving the accuracy and specificity of catalyst formula optimization.

[0075] Furthermore, we randomly configured several platinum group nanomaterial catalyst formulas, combined with preparation condition parameters, indexed the catalyst preparation sample set, and obtained several high-frequency catalyst structures, including:

[0076] S21, extracting a first platinum group nanomaterial catalyst formula from the plurality of platinum group nanomaterial catalyst formulas;

[0077] S22, retrieving historical preparation experimental data that meets the first platinum group nanomaterial catalyst formula and the preparation condition parameters, wherein the historical preparation experimental data includes a plurality of catalyst structure record data;

[0078] S23, performing structural similarity clustering on the plurality of catalyst structure record data to obtain multiple clusters of catalyst structure record data;

[0079] S24. Based on the multiple clusters of catalyst structure record data, clusters whose intra-cluster size is less than or equal to the intra-cluster size threshold are deleted to obtain the multiple high-frequency catalyst structures, where the intra-cluster size is equal to the amount of catalyst structure record data within the cluster.

[0080] In a preferred embodiment, first, a number of randomly configured platinum group nanomaterial catalyst formulas are extracted in sequence, and one platinum group nanomaterial catalyst formula is extracted each time as a first platinum group nanomaterial catalyst formula. Then, historical preparation experimental data that meets the first platinum group nanomaterial catalyst formula and preparation condition parameters are retrieved, thereby fully utilizing the information value of historical experimental records. By accurately matching the formula composition and preparation conditions of the first platinum group nanomaterial catalyst formula, relevant historical experimental records are retrieved from the historical database as historical preparation experimental data. These historical preparation experimental data include multiple catalyst structure record data, each catalyst structure record data contains catalyst microstructure information formed under specific formula and preparation conditions, such as crystal phase composition, morphology characteristics, particle size distribution, etc., reflecting the actual situation that the same formula and preparation parameters may produce multiple different structures.

[0081] Subsequently, the multiple catalyst structure record data retrieved are clustered for structural similarity. Specifically, a structural similarity algorithm is used to calculate the similarity between different structure records, and then cluster analysis is performed based on the similarity, and records with similar structural features are classified into the same cluster. Through this clustering process, it is possible to identify several main structural types that may be formed under given formulations and preparation conditions, and obtain multi-cluster catalyst structure record data, each cluster representing a class of catalysts with similar structures. Among them, the calculation of structural similarity can adopt a cosine similarity method based on eigenvectors, that is, each catalyst structure is represented as a multidimensional eigenvector, including key features such as crystal phase composition ratio, average particle size, surface coordination number, and electronic structure parameters, and then the cosine value of the angle between the two eigenvectors is calculated as a similarity index. The closer the similarity value is to 1, the more similar the two structures are. In addition, the structural similarity can also be calculated using a graph theory method, where the catalyst structure is represented as a topological graph, and the structural similarity is quantified by comparing the size of the adjacency matrix or the largest common subgraph of the graph. Based on the calculated similarity matrix, K-means or hierarchical clustering algorithms are used to complete structural classification, forming multiple catalyst structure clusters with different structural characteristics, and obtaining multi-cluster catalyst structure record data.

[0082] Then, based on the multi-cluster catalyst structure record data, clusters with an intra-cluster size less than or equal to the intra-cluster size threshold are deleted to obtain several high-frequency catalyst structures. The intra-cluster size refers to the amount of catalyst structure record data contained in the cluster, reflecting the frequency of occurrence of this type of structure. By setting the intra-cluster size threshold, catalyst structure types with high frequency and good repeatability are screened out, and low-frequency structures that may have formed accidentally or due to experimental error are filtered out. This statistical frequency-based screening mechanism ensures that subsequent performance evaluation and optimization are based on the most representative and repeatable catalyst structures.

[0083] It's worth noting that for a given platinum-group nanomaterial catalyst formulation, multiple high-frequency catalyst structures may meet the cluster size requirements. In this case, the fitness values ​​of all high-frequency catalyst structures are calculated, and the average is taken as the overall fitness value for the formulation. This approach fully accounts for the structural diversity and uncertainty in the catalyst preparation process, making the fitness evaluation more comprehensive and robust.

[0084] Through the above steps, an effective mapping from catalyst formula to high-frequency structure is achieved, which solves the uncertainty problem between theoretical formula and actual structure in catalyst research and development, provides a reliable structural foundation for subsequent performance prediction and optimization, and improves the practical application value and reliability of the optimization results.

[0085] Furthermore, the steps for constructing the catalyst performance mapping model include:

[0086] S31, collecting platinum group nanomaterial catalyst structure data and labels identifying catalyst performance data, and training a first weak learner;

[0087] S32. Counting the first residual vector of the first weak learner;

[0088] S33. When the modulus of the first residual vector is greater than or equal to the convergence modulus, update the supervision data with the first residual vector and train the second weak learner;

[0089] S34 , until the modulus of the Mth residual vector is less than the convergence modulus, summing and integrating the outputs of the first weak learner to the Mth weak learner to obtain the catalyst performance mapping model.

[0090] In a preferred embodiment, an integrated learning catalyst performance mapping model is constructed. By gradually training multiple weak learners and integrating their outputs, the complex nonlinear relationship between catalyst structure and performance can be effectively captured, providing reliable performance prediction support for formulation optimization.

[0091] First, the structural data of platinum-group nanomaterial catalysts and labels identifying catalyst performance data are collected to train the first weak learner. Specifically, the structural data of platinum-group nanomaterial catalysts include the microstructural characteristics of the catalyst, such as crystal phase composition, morphological characteristics, particle size distribution, and electronic structure parameters; while the labels of the catalyst performance data correspond to the measured values ​​of various preset performance indicators, such as catalytic activity and stability. Based on these structure-performance correspondence data, the first weak learner is trained to establish a preliminary structure-performance mapping relationship. Among them, the weak learner can be a decision tree, neural network, or other machine learning model. Although its predictive ability is limited, it can form a powerful overall predictive ability through subsequent integration.

[0092] Then, the first residual vector of the first weak learner is calculated. This first residual vector reflects the difference between the first weak learner's predicted value and the actual label value and serves as an indicator for evaluating the accuracy of the first residual vector. Specifically, by calculating the difference between the predicted catalyst performance data and the label identifying the catalyst performance data when the first weak learner processes the platinum-group nanomaterial catalyst structure data, the first residual vector is formed. This represents the information that the first weak learner failed to capture.

[0093] When the modulus of the first residual vector is greater than or equal to the convergence modulus, the first residual vector is used to update the supervisory data and train the second weak learner. Specifically, a determination is made as to whether the current residual vector meets the preset convergence criteria, i.e., whether the modulus of the residual vector is greater than or equal to the convergence modulus. If the modulus of the residual vector is still large, indicating that the model accuracy needs to be improved, the first residual vector is used as a new supervisory signal to train the second weak learner, focusing on learning the parts that the first weak learner failed to accurately capture. This residual-based iterative training mechanism can gradually improve the overall prediction accuracy of the model.

[0094] The above iterative process is repeated until the modulus of the Mth residual vector is less than the convergence modulus. The outputs of the first through Mth weak learners are then summed and integrated to obtain a catalyst performance mapping model. Iterative training terminates when the modulus of the residual vector is sufficiently small to meet the preset convergence criterion (i.e., when the modulus of the residual vector is less than the convergence modulus). The outputs of all weak learners are weighted and summed to form the final catalyst performance mapping model. This integrated model combines the predictive capabilities of multiple weak learners and can more accurately capture the complex relationship between the structure and performance of platinum-group nanomaterial catalysts.

[0095] Through the above steps, a catalyst performance mapping model with high precision and strong generalization ability was constructed, which can accurately predict its various performance indicators based on the catalyst structure information, providing a reliable evaluation tool for the formulation optimization of platinum group nanomaterial catalysts, and improving the optimization efficiency and accuracy.

[0096] Furthermore, performing optimization based on the plurality of fitnesses and the plurality of platinum group nanomaterial catalyst formulations to determine a target catalyst formulation includes:

[0097] S41. Construct gravity calculation formula:

[0098] ,

[0099] in, Characterize the zth platinum group nanomaterial catalyst formulation, Characterize the kth platinum group nanomaterial catalyst formulation, Characterizes the gravitational constant, Characterize the fitness value of the zth platinum group nanomaterial catalyst formulation, Characterize the fitness value of the kth platinum group nanomaterial catalyst formulation, characterizing the attractiveness of the zth platinum group nanomaterial catalyst formulation and the kth platinum group nanomaterial catalyst formulation, Characterize the Euclidean distance between the zth platinum group nanomaterial catalyst formula and the kth platinum group nanomaterial catalyst formula, Characterize the direction vector, pointing to the zth platinum group nanomaterial catalyst formula, Characterizes a small constant, Characterizes the rate of decrease of a predefined gravitational constant, Characterizes the preset number of convergence iterations, Indicates the current number of iterations;

[0100] S42. Based on the gravity calculation formula, the plurality of fitnesses, and the plurality of platinum group nanomaterial catalyst formulas, executing a formula update at a preset movement speed to obtain an updated platinum group nanomaterial catalyst formula;

[0101] S43. Perform iterative update according to the updated platinum group nanomaterial catalyst formula until the preset convergence iteration number is met to obtain the target catalyst formula.

[0102] In a preferred embodiment, during the optimization process of the platinum group nanomaterial catalyst formulation, an optimization algorithm based on a gravitational search mechanism is designed, which can efficiently find the optimal solution in the multi-objective optimization space of formulation performance, thereby realizing intelligent search and iterative optimization of the formulation space.

[0103] First, construct the gravity calculation formula as In this formula, Characterize the zth platinum group nanomaterial catalyst formulation, Characterize the kth platinum group nanomaterial catalyst formulation, which represents different points in the multi-objective optimization space of formulation performance; Characterizes the gravitational constant, which controls the overall strength of gravitational attraction; Characterizes the fitness value of the zth platinum group nanomaterial catalyst formula, and f(xk) characterizes the fitness value of the kth platinum group nanomaterial catalyst formula. These two fitness values ​​determine the quality of the platinum group nanomaterial catalyst formula. The higher the fitness, the stronger the gravitational effect. Characterizes the attraction between the zth platinum group nanomaterial catalyst formulation and the kth platinum group nanomaterial catalyst formulation, which is a vector representation of the force between the two formulations; The Euclidean distance between the zth platinum group nanomaterial catalyst formula and the kth platinum group nanomaterial catalyst formula reflects the distance between the two points in the formula space; Characterizing the direction vector, pointing to the zth platinum group nanomaterial catalyst formula, adjusting the kth platinum group nanomaterial catalyst formula based on the zth platinum group nanomaterial catalyst formula, and ensuring the directionality and effectiveness of the gravitational effect; Characterizes a small constant used to avoid computational singularities when the distance is zero; represents the rate of decrease of the predefined gravitational constant; T represents the preset number of convergence iterations, and t represents the current number of iterations. These two parameters jointly control the transition process from global exploration to local development.

[0104] The term controls the balance between global exploration and local exploitation. The parameter is a factor that adjusts the conversion rate between global exploration and local development. When the value is high, decays faster, resulting in a gravitational constant It decreases rapidly in the early iteration stage, which weakens the interaction between different formulations and shifts from a large-scale global search to a detailed development of potential areas in advance. It is suitable for situations where the dimensions of the catalyst formulation space are small or the optimal solution distribution area has been roughly determined by prior knowledge. However, too high There is a potential risk in using a value of , which may lead to premature focus on local areas and ignore other high-performance areas that may exist in the recipe space, thus missing the global optimal solution. When the value is low, It slowly decays with the increase of the number of iterations, so that the global exploration ability is maintained in a long iteration cycle. The movement of each recipe point in the space is more active, and different areas of the recipe space can be explored more comprehensively. It is suitable for the situation where there are many dimensions in the catalyst recipe space, where the optimal recipe may be distributed in multiple different component ratio areas, and sufficient exploration ability is required to traverse these areas. However, low The cost of the value strategy is that the convergence speed is slow and more iterations are required to reach convergence, so the computing resources are consumed more. The value is set in the range of 10-20 and fine-tuned according to the specific catalyst type and application scenario to achieve the best balance between global exploration and local exploitation, ensuring high-quality optimization results at an acceptable computational cost.

[0105] Subsequently, based on the constructed gravity calculation formula, using several fitnesses and several platinum group nanomaterial catalyst formulas, the formula update is performed according to the preset moving speed to obtain an updated platinum group nanomaterial catalyst formula. Specifically, the gravitational force of each formula on all other formulas is calculated, and then the position coordinates of the formula are updated according to this gravitational force and the preset moving speed to achieve iterative optimization of the formula, so that the formula with high fitness has a stronger attraction to other formulas, guiding the search to move to the high fitness area, and obtaining an updated platinum group nanomaterial catalyst formula. Afterwards, the formula update process of step S42 is repeated continuously. In each iteration, The value will gradually decrease, causing the algorithm to gradually shift from global exploration to local optimization. When the preset number of iterations is reached or other convergence conditions are met, the recipe with the highest fitness in the recipe set is selected as the target catalyst recipe.

[0106] Through the above steps, efficient optimization of the platinum group nanomaterial catalyst formula was achieved, and the optimal solution could be quickly found in the complex multi-dimensional formula space, providing strong technical support for catalyst research and development.

[0107] Example 2, as Figure 2 As shown, based on the same inventive concept as the platinum group nanomaterial catalyst formulation optimization method provided in Example 1, this embodiment of the present invention further provides a platinum group nanomaterial catalyst formulation optimization system, comprising:

[0108] An optimization parameter construction module 11 is used to construct a multi-objective optimization space for recipe performance based on predefined performance indicator boundary values, and to weight preset performance indicators based on predefined performance indicator weight distribution to construct a fitness function belonging to the multi-objective optimization space for recipe performance, wherein the preset performance indicators are positive indicators;

[0109] A recipe generation module 12 is used to randomly configure a number of platinum group nanomaterial catalyst recipes, combine preparation condition parameters, index the catalyst preparation sample set, and obtain a number of high-frequency catalyst structures;

[0110] The fitness calculation module 13 is used to process the plurality of high-frequency catalyst structures respectively through a pre-trained catalyst performance mapping model to obtain a plurality of performance index prediction values, input the fitness function, and obtain a plurality of fitness values;

[0111] The optimization output module 14 is used to perform optimization based on the multiple fitness values ​​and the multiple platinum group nanomaterial catalyst formulas, determine the target catalyst formula, and send it to the catalyst formula management and control end.

[0112] Furthermore, the optimization parameter construction module 11 includes the following execution steps:

[0113] receiving a target application scenario, performing a correlation analysis on preset performance indicators of the platinum group nanomaterial catalyst, and obtaining a weight distribution of the performance indicators;

[0114] A catalyst sample set of the target application scenario is statistically analyzed, and a mode analysis is performed on the preset performance indicators of the platinum group nanomaterial catalyst to obtain the boundary values ​​of the performance indicators.

[0115] Furthermore, the optimization parameter construction module 11 further includes the following execution steps:

[0116] Extracting user-preconfigured evaluation trait attributes from the target application scenario;

[0117] Based on the target application scenario, collecting one-to-one correspondences of: evaluation trait attribute characteristic value sequence, first preset performance indicator characteristic value sequence, up to Nth preset performance indicator characteristic value sequence;

[0118] Using the evaluation trait attribute characteristic value sequence as a reference sequence and the first preset performance indicator characteristic value sequence to the Nth preset performance indicator characteristic value sequence as a comparison sequence, performing grey relational analysis to generate first preset performance indicator correlation degrees to Nth preset performance indicator correlation degrees;

[0119] The first preset performance indicator correlation degree is traversed until the Nth preset performance indicator correlation degree, and the correlation degrees are summed and compared to obtain the performance indicator weight distribution.

[0120] Furthermore, the optimization parameter construction module 11 further includes the following execution steps:

[0121] Collecting a catalyst sample set with target product quality that meets the target application scenario, wherein the target product quality refers to the substance that needs to be produced by the catalyst;

[0122] Extracting a first set of preset performance indicator recorded values ​​based on the catalyst sample set, performing box plot analysis to obtain a first set of preset performance indicator recorded values, and constructing a first preset performance indicator boundary value based on the maximum and minimum values ​​of the first set of preset performance indicator recorded values;

[0123] Until the Nth preset performance indicator record value set is extracted based on the catalyst sample set, a box plot analysis is performed to obtain the preset performance indicator record value in the Nth set, and the Nth preset performance indicator boundary value is constructed with the maximum and minimum values ​​of the preset performance indicator record values ​​in the Nth set.

[0124] Furthermore, the mathematical model of the fitness function is as follows:

[0125] ,

[0126] ,

[0127] in, Characterize the fitness value of any platinum group nanomaterial catalyst formula, Characterize the predicted value of the performance index of platinum group nanomaterial catalyst formula, x represents any platinum group nanomaterial catalyst formula, Characterizes the minimum boundary value of the preset performance indicator of the i-th attribute, Characterizes the maximum boundary value of the preset performance indicator of the i-th attribute, Characterizes the preset performance indicator weight of the i-th attribute.

[0128] Furthermore, the recipe generation module 12 includes the following execution steps:

[0129] extracting a first platinum group nanomaterial catalyst formula from the plurality of platinum group nanomaterial catalyst formulas;

[0130] Retrieving historical preparation experimental data that meets the first platinum group nanomaterial catalyst formula and the preparation condition parameters, wherein the historical preparation experimental data includes a plurality of catalyst structure record data;

[0131] performing structural similarity clustering on the plurality of catalyst structure record data to obtain a plurality of clusters of catalyst structure record data;

[0132] Based on the multiple clusters of catalyst structure record data, clusters with an intra-cluster size less than or equal to an intra-cluster size threshold are deleted to obtain the multiple high-frequency catalyst structures, where the intra-cluster size is equal to the amount of catalyst structure record data within the cluster.

[0133] Furthermore, the steps for constructing the catalyst performance mapping model include:

[0134] Collect platinum group nanomaterial catalyst structure data and labels identifying catalyst performance data to train the first weak learner;

[0135] Counting the first residual vector of the first weak learner;

[0136] When the modulus of the first residual vector is greater than or equal to the convergence modulus, the supervision data is updated with the first residual vector to train the second weak learner;

[0137] Until the modulus of the Mth residual vector is less than the convergence modulus, the outputs of the first weak learner to the Mth weak learner are summed and integrated to obtain the catalyst performance mapping model.

[0138] Furthermore, the optimization output module 14 includes the following execution steps:

[0139] Construct the gravity calculation formula:

[0140] ,

[0141] in, Characterize the zth platinum group nanomaterial catalyst formulation, Characterize the kth platinum group nanomaterial catalyst formulation, Characterizes the gravitational constant, Characterize the fitness value of the zth platinum group nanomaterial catalyst formulation, Characterize the fitness value of the kth platinum group nanomaterial catalyst formulation, characterizing the attractiveness of the zth platinum group nanomaterial catalyst formulation and the kth platinum group nanomaterial catalyst formulation, Characterize the Euclidean distance between the zth platinum group nanomaterial catalyst formula and the kth platinum group nanomaterial catalyst formula, Characterize the direction vector, pointing to the zth platinum group nanomaterial catalyst formula, Characterizes a small constant, Characterizes the rate of decrease of a predefined gravitational constant, Characterizes the preset number of convergence iterations, Indicates the current number of iterations;

[0142] Based on the gravity calculation formula, based on the several fitnesses and the several platinum group nanomaterial catalyst formulas, executing formula update at a preset moving speed to obtain an updated platinum group nanomaterial catalyst formula;

[0143] Iterative updating is performed according to the updated platinum group nanomaterial catalyst formula until the preset convergence iteration number is met to obtain the target catalyst formula.

[0144] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0147] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0149] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0150] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for optimizing the formulation of platinum group nanomaterial catalysts, characterized in that: include: According to the predefined performance indicator boundary values, a multi-objective optimization space for recipe performance is constructed, and based on the predefined performance indicator weight distribution, the preset performance indicators are weighted to construct a fitness function belonging to the multi-objective optimization space for recipe performance, where the preset performance indicators are positive indicators; Randomly configure several platinum group nanomaterial catalyst formulas, combine preparation condition parameters, index the catalyst preparation sample set, and obtain several high-frequency catalyst structures; Processing the plurality of high-frequency catalyst structures respectively through a pre-trained catalyst performance mapping model to obtain a plurality of performance index prediction values, and inputting the plurality of performance index prediction values ​​into the fitness function to obtain a plurality of fitness values; Performing optimization based on the plurality of fitnesses and the plurality of platinum group nanomaterial catalyst formulas, determining a target catalyst formula, and sending the result to a catalyst formula control terminal; Among them, several platinum group nanomaterial catalyst formulas were randomly configured, combined with preparation condition parameters, and the catalyst preparation sample set was indexed to obtain several high-frequency catalyst structures, including: extracting a first platinum group nanomaterial catalyst formula from the plurality of platinum group nanomaterial catalyst formulas; Retrieving historical preparation experimental data that meets the first platinum group nanomaterial catalyst formula and the preparation condition parameters, wherein the historical preparation experimental data includes a plurality of catalyst structure record data; performing structural similarity clustering on the plurality of catalyst structure record data to obtain a plurality of clusters of catalyst structure record data; Based on the multiple clusters of catalyst structure record data, clusters with an intra-cluster size less than or equal to an intra-cluster size threshold are deleted to obtain the multiple high-frequency catalyst structures, where the intra-cluster size is equal to the amount of catalyst structure record data within the cluster.

2. The method according to claim 1, wherein According to the predefined performance indicator boundary values, a multi-objective optimization space for recipe performance is constructed. Based on the predefined performance indicator weight distribution, the preset performance indicators are weighted to construct a fitness function belonging to the multi-objective optimization space for recipe performance, which includes: receiving a target application scenario, performing a correlation analysis on preset performance indicators of the platinum group nanomaterial catalyst, and obtaining a weight distribution of the performance indicators; A catalyst sample set of the target application scenario is statistically analyzed, and a mode analysis is performed on the preset performance indicators of the platinum group nanomaterial catalyst to obtain the boundary values ​​of the performance indicators.

3. The method according to claim 2, wherein Receive the target application scenario, perform correlation analysis on the preset performance indicators of the platinum group nanomaterial catalyst, and obtain the weight distribution of the performance indicators, including: Extracting user-preconfigured evaluation trait attributes from the target application scenario; Based on the target application scenario, collecting one-to-one correspondences of: evaluation trait attribute characteristic value sequence, first preset performance indicator characteristic value sequence, up to Nth preset performance indicator characteristic value sequence; Using the evaluation trait attribute characteristic value sequence as a reference sequence and the first preset performance indicator characteristic value sequence to the Nth preset performance indicator characteristic value sequence as a comparison sequence, performing grey relational analysis to generate first preset performance indicator correlation degrees to Nth preset performance indicator correlation degrees; The first preset performance indicator correlation degree is traversed until the Nth preset performance indicator correlation degree, and the correlation degrees are summed and compared to obtain the performance indicator weight distribution.

4. The method according to claim 2, wherein Collect a sample set of catalysts for the target application scenario, perform mode analysis on the preset performance indicators of the platinum group nanomaterial catalyst, and obtain the boundary values ​​of the performance indicators, including: Collecting a catalyst sample set with target product quality that meets the target application scenario, wherein the target product quality refers to the substance that needs to be produced by the catalyst; Extracting a first set of preset performance indicator recorded values ​​based on the catalyst sample set, performing box plot analysis to obtain a first set of preset performance indicator recorded values, and constructing a first preset performance indicator boundary value based on the maximum and minimum values ​​of the first set of preset performance indicator recorded values; Until the Nth preset performance indicator record value set is extracted based on the catalyst sample set, a box plot analysis is performed to obtain the preset performance indicator record value in the Nth set, and the Nth preset performance indicator boundary value is constructed with the maximum and minimum values ​​of the preset performance indicator record values ​​in the Nth set.

5. The method according to claim 1, wherein The mathematical model of the fitness function is as follows: ， ， in, Characterize the fitness value of any platinum group nanomaterial catalyst formula, Characterize the predicted value of the performance index of platinum group nanomaterial catalyst formula, x represents any platinum group nanomaterial catalyst formula, Characterizes the minimum boundary value of the preset performance indicator of the i-th attribute, Characterizes the maximum boundary value of the preset performance indicator of the i-th attribute, Characterizes the preset performance indicator weight of the i-th attribute.

6. The method according to claim 1, wherein The steps to construct the catalyst performance mapping model include: Collect platinum group nanomaterial catalyst structure data and labels identifying catalyst performance data to train the first weak learner; Counting the first residual vector of the first weak learner; When the modulus of the first residual vector is greater than or equal to the convergence modulus, the supervision data is updated with the first residual vector to train the second weak learner; Until the modulus of the Mth residual vector is less than the convergence modulus, the outputs of the first weak learner to the Mth weak learner are summed and integrated to obtain the catalyst performance mapping model.

7. The method according to claim 1, wherein Performing optimization based on the plurality of fitnesses and the plurality of platinum group nanomaterial catalyst formulations to determine a target catalyst formulation includes: Construct the gravity calculation formula: ， in, Characterize the zth platinum group nanomaterial catalyst formulation, Characterize the kth platinum group nanomaterial catalyst formulation, Characterizes the gravitational constant, Characterize the fitness value of the zth platinum group nanomaterial catalyst formulation, Characterize the fitness value of the kth platinum group nanomaterial catalyst formulation, characterizing the attractiveness of the zth platinum group nanomaterial catalyst formulation and the kth platinum group nanomaterial catalyst formulation, Characterize the Euclidean distance between the zth platinum group nanomaterial catalyst formula and the kth platinum group nanomaterial catalyst formula, Characterize the direction vector, pointing to the zth platinum group nanomaterial catalyst formula, Characterizes a small constant, Characterizes the rate of decrease of a predefined gravitational constant, Characterizes the preset number of convergence iterations, Indicates the current number of iterations; Based on the gravity calculation formula, based on the several fitnesses and the several platinum group nanomaterial catalyst formulas, executing formula update at a preset moving speed to obtain an updated platinum group nanomaterial catalyst formula; Iterative updating is performed according to the updated platinum group nanomaterial catalyst formula until the preset convergence iteration number is met to obtain the target catalyst formula.

8. A platinum group nanomaterial catalyst formulation optimization system, characterized in that: For implementing the method according to any one of claims 1 to 7, comprising: An optimization parameter construction module is used to construct a multi-objective optimization space for recipe performance based on predefined performance indicator boundary values, and to weight preset performance indicators based on predefined performance indicator weight distribution to construct a fitness function belonging to the multi-objective optimization space for recipe performance, wherein the preset performance indicators are positive indicators; The recipe generation module is used to randomly configure several platinum group nanomaterial catalyst recipes, combine preparation condition parameters, index the catalyst preparation sample set, and obtain several high-frequency catalyst structures; a fitness calculation module, configured to process the plurality of high-frequency catalyst structures respectively using a pre-trained catalyst performance mapping model to obtain a plurality of performance index prediction values, and input the plurality of performance index prediction values ​​into the fitness function to obtain a plurality of fitness values; The optimization output module is used to perform optimization based on the multiple fitness values ​​and the multiple platinum group nanomaterial catalyst formulas, determine the target catalyst formula, and send it to the catalyst formula management and control end.

Images