Molecular sieve type prediction method and device
Through the molecular sieve type prediction method based on the factor decomposition machine, the problem of difficult product type prediction in the molecular sieve catalyst synthesis experiment was solved, and the effect of high accuracy prediction and experimental cost reduction was achieved.
Patent Information
- Application Number
- CN202311618262.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-29
- Publication Date
- 2025-05-30
AI Technical Summary
In the field of petrochemical industry, the impact of changes in experimental operating conditions of molecular sieve catalysts on the type of product molecular sieve is difficult to predict, resulting in increased experimental costs and inefficient research and development of new molecular sieves.
The molecular sieve type prediction method based on a factor decomposition machine is used to predict the product molecular sieve type by obtaining the experimental operating conditions of molecular sieve synthesis and constructing a prediction model. The method includes data preprocessing, factor analysis and model training, and optimizes model losses using stochastic gradient descent.
It realizes the high accuracy of predicting the type of product molecular sieve based on the operating conditions of the synthetic experiment, guiding actual synthetic experiments, reducing unnecessary experiments, and reducing R&D costs.
Smart Images
Figure CN120072077A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of molecular sieve catalyst processes, and particularly to a method and device for predicting the type of molecular sieve. Background Art
[0002] In the field of petrochemical engineering, molecular sieve catalysts have become one of the most widely used catalyst materials. During the molecular sieve synthesis experiment process, the influence of changes in experimental operating conditions on the resulting molecular sieve can only be verified through specific molecular sieve synthesis experiments. This has increased the experimental cost and hindered the efficiency of the research and development of new molecular sieves. Therefore, there is an urgent need for a prediction method to predict the corresponding product molecular sieve type based on different experimental operating conditions, providing guidance for the actual molecular sieve synthesis experiment process. Summary of the Invention
[0003] The purpose of the embodiments of this application is to provide a method and device for predicting the type of molecular sieve.
[0004] To achieve the above purpose, the first aspect of this application provides a method for predicting the type of molecular sieve, including: obtaining molecular sieve synthesis experiment operating condition data; and inputting the molecular sieve synthesis experiment operating condition data into a molecular sieve type prediction model based on a factorization machine to obtain product molecular sieve type data corresponding to the molecular sieve synthesis experiment operating condition data.
[0005] In the embodiments of this application, the molecular sieve type prediction model based on a factorization machine is constructed through the following operations: obtaining molecular sieve synthesis experiment data and constructing a molecular sieve synthesis experiment data set, where the molecular sieve synthesis experiment data includes molecular sieve synthesis experiment operating condition data and corresponding product molecular sieve type data; based on this data set, obtaining unique operating conditions and experimental data on whether the corresponding molecular sieve is successfully synthesized; based on this experimental data, performing factor analysis on the influencing factor data, constructing structured data for model training and establishing a data set; and based on the data set, constructing and training a molecular sieve type prediction model based on a factorization machine.
[0006] In the embodiments of this application, before obtaining unique operating conditions and experimental data on whether the corresponding molecular sieve is successfully synthesized, it further includes: preprocessing the molecular sieve synthesis experiment data in the data set, and this preprocessing includes missing value processing and outlier processing. Among them, the missing value processing includes: pairing the molecular sieve experiment operating condition data and the product molecular sieve type data one by one, and removing the sample data missing one of them; among them, the outlier processing includes: discarding the abnormal data samples, and the abnormal data samples include the sample data where the temperature or pressure deviates from the set value due to the abnormality of the reaction instrument.
[0007] In an embodiment of the present application, obtaining experimental data with unique operating conditions and whether the corresponding molecular sieve is successfully synthesized includes: traversing the data set to compare whether the operating conditions of different sample data are consistent, and if so, clustering the sample data into one category; traversing all categories containing multiple samples to determine whether the molecular sieve types in the same category are consistent, and if so, using the type as the molecular sieve type corresponding to the sample data of this category, and if not, discarding the sample data of this category.
[0008] In an embodiment of the present application, based on the experimental data, factor analysis is performed on the influencing factor data, and structured data for model training is constructed and a data set is established, including: using one-hot encoding to encode the success or failure of the synthetic molecular sieve type in the experimental data into a 0-1 form as a label for the model output; using the molecular sieve synthesis experimental operating condition data in the experimental data as input data, performing correlation calculation on the input data, and evaluating the KMO value and Bartlett sphericity; and eliminating the experimental data whose KMO value and Bartlett sphericity do not meet the preset standards, and using the remaining data as structured data for model training, using part of it as a training set, and using the other part as a validation set.
[0009] In an embodiment of the present application, constructing and training a molecular sieve type prediction model based on a factor decomposition machine includes: establishing a molecular sieve type prediction model based on a factor decomposition machine, optimizing the network loss Loss of the molecular sieve type prediction model using a stochastic gradient descent method; and training the molecular sieve type prediction model, when the training set has trained the molecular sieve type prediction model for a preset number of times, performing an error test on the validation set, stopping the training when the error of the validation set increases, and saving the molecular sieve type prediction model.
[0010] In the embodiment of the present application, the hidden vector dimension of the molecular sieve type prediction model is uniformly set to 128.
[0011] In the embodiment of the present application, the network loss Loss is a cross entropy loss.
[0012] In the embodiments of the present application, the molecular sieve synthesis experimental operating conditions data include one or more of the following: the type and content of raw materials for molecular sieve synthesis, reaction temperature, reaction time and stirring data; the product molecular sieve type data include one or more of the following: ZSM-5, SAPO, MOR and Beta types.
[0013] A second aspect of the present application provides a device for predicting the type of molecular sieve, the device comprising: a memory; and a processor, the processor being configured to execute the above-mentioned molecular sieve type prediction method.
[0014] A third aspect of the present application provides a machine-readable storage medium, on which instructions are stored. It is characterized in that when the instructions are executed by a processor, the processor is configured to execute the above-mentioned molecular sieve type prediction method.
[0015] A fourth aspect of the present application provides a computer program product, including a computer program, which when executed by a processor, implements the above-mentioned molecular sieve type prediction method.
[0016] Through the above technical solutions, the present invention only needs the synthetic experimental data stored historically to build a model, and can accurately predict the type of the product molecular sieve based on the synthetic experimental operation conditions, providing guidance for the actual molecular sieve synthesis experiment, so as to adjust the experimental operation conditions targeted to avoid unnecessary experiments and reduce the R & D experimental cost.
[0017] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation part. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. They are used to explain the embodiments of the present application together with the following specific implementation manners, but do not constitute a limitation to the embodiments of the present application. In the drawings:
[0019] Figure 1 Schematically shows a flow diagram of the molecular sieve type prediction method according to an embodiment of the present application.
[0020] Figure 2 Schematically shows a flow diagram of building a molecular sieve type prediction model according to an embodiment of the present application.
[0021] Figure 3 Schematically shows a schematic diagram of the change of the loss curve during the training process of the molecular sieve type prediction method based on the factorization machine according to an embodiment of the present application.
[0022] Figure 4 Schematically shows a flow diagram of a molecular sieve type prediction method based on the factorization machine according to an embodiment of the present application.
[0023] Figure 5 Schematically shows an internal structure diagram of the device for predicting the molecular sieve type according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will, with reference to the accompanying drawings in the embodiments of this application, clearly and completely describe the technical solutions in the embodiments of this application. It should be understood that the specific embodiments described herein are only for explaining and illustrating the embodiments of this application, and are not used to limit the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts fall within the scope of protection of this application.
[0025] It should be noted that if there are directional indications (such as up, down, left, right, front, back...) involved in the embodiments of this application, then such directional indications are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If this specific posture changes, then such directional indications will also change accordingly.
[0026] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of this application, then such descriptions of "first", "second", etc. are only for descriptive purposes, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.
[0027] Figure 1 Schematically shown is a flowchart of a molecular sieve type prediction method according to an embodiment of this application. As Figure 1 shown, in an embodiment of this application, a molecular sieve type prediction method is provided, including the following steps:
[0028] Step 102, obtaining molecular sieve synthesis experimental operation condition data; and
[0029] Step 104, inputting the molecular sieve synthesis experimental operation condition data into a molecular sieve type prediction model based on a factorization machine to obtain product molecular sieve type data corresponding to the molecular sieve synthesis experimental operation condition data.
[0030] The molecular sieve type prediction model is as follows:
[0031]
[0032] where x i and x j represent the i-th and j-th features of each sample, and w iis the learning weight of the i-th feature, w 0 is the initial weight. n represents that each sample has n features, and these n features constitute the molecular sieve synthesis experimental operation condition data, v i is the latent vector of the i-th dimensional feature, v j is the latent vector of the j-th dimensional feature. <,> represents the dot product of vectors. The length of the latent vector is k (k << n), and it includes k factors describing the features. Here, the feature refers to multiple input data of each sample, that is, the molecular sieve synthesis experimental operation condition data. For example, the i-th and j-th features can respectively refer to the reaction temperature and reaction time in the molecular sieve synthesis experimental operation condition data.
[0033] In the embodiments of the present application, the molecular sieve synthesis experimental operation condition data may include one or more of the following: the types and contents of raw materials for molecular sieve synthesis, reaction temperature, reaction time, and stirring data. For example, it may include all of them, a total of 68-dimensional data; the product molecular sieve type data may include one or more of the following: ZSM-5, SAPO, MOR, and Beta types. For example, it may include all of them, a total of 4-dimensional data.
[0034] Through the above technical solution, it is possible to accurately predict the type of product molecular sieve based on the synthesis experimental operation conditions, provide guidance for actual molecular sieve synthesis experiments, and thus targetedly adjust the experimental operation conditions to avoid unnecessary experiments and reduce the R & D experimental costs.
[0035] Figure 2 Schematically shows a flow chart of constructing a molecular sieve type prediction model according to an embodiment of the present application. The following combines Figure 2 , to introduce the construction process of the above molecular sieve type prediction model based on the factorization machine.
[0036] In the embodiments of the present application, the molecular sieve type prediction model based on the factorization machine is constructed through the following operations:
[0037] Step 202, obtain molecular sieve synthesis experimental data, and construct a molecular sieve synthesis experimental data set. The molecular sieve synthesis experimental data includes molecular sieve synthesis experimental operation condition data and corresponding product molecular sieve type data.
[0038] Specifically, in step 202, operating condition data and molecular sieve type data of a series of experimental batches are obtained as the molecular sieve synthesis experimental data set. That is, the molecular sieve synthesis experimental data set includes the operating condition data of the molecular sieve synthesis experiment and the corresponding product molecular sieve type data. The operating condition data of the molecular sieve synthesis experiment includes the types and contents of raw materials for molecular sieve synthesis, reaction temperature, reaction time, and stirring data, with a total of 68-dimensional data; the molecular sieve type data includes ZSM-5, SAPO, MOR, and Beta types, with a total of 4-dimensional data.
[0039] The sample data in the molecular sieve synthesis experimental data set can be further processed to obtain a more reasonable new data set. The preprocessing includes missing value processing and outlier processing; the missing value processing is specifically as follows: in the molecular sieve synthesis experimental data set, first, the operating condition data of the molecular sieve experiment and the molecular sieve type data are paired one by one, and the sample data missing any one of them is removed; the outlier processing is specifically to discard the obviously abnormal data samples, including the sample data with the temperature or pressure deviating from the set value due to the abnormal reaction instrument.
[0040] Step 204, based on the data set, obtain the unique operating conditions and experimental data on whether the corresponding molecular sieve is successfully synthesized.
[0041] Specifically, the input data under different operating conditions can be sorted out, and the entire data set can be traversed to obtain the unique operating conditions and the corresponding product molecular sieve types to ensure the uniqueness of the operating condition conditions of each sample in the data set. The unique operating conditions and the corresponding product molecular sieve types mean that one operating condition corresponds to one molecular sieve type.
[0042] Further, step 204 specifically includes the following steps:
[0043] Traverse the entire preprocessed data set, compare whether the operating conditions of different sample data are the same. If so, cluster the sample data into one category;
[0044] To ensure that one operating condition corresponds to one molecular sieve type, traverse all categories in the data set that contain multiple samples, and judge whether the molecular sieve types in the same category are the same. If so, use this type as the molecular sieve type corresponding to the sample data of this category. If not, discard the sample data of this category.
[0045] Step 206, based on this experimental data, perform factor analysis on the influencing factor data, construct structured data for model training, and establish a data set.
[0046] Specifically, this step takes data of different operating conditions as input data, aligns them with the molecular sieve category data of the output data, constructs structured data for model training, and divides the entire data set into a training set and a validation set.
[0047] Specifically, this step may include the following operations:
[0048] Step a: Count the molecular sieves that are successfully synthesized in the statistical data set, and use one-hot coding for the molecular sieve type data. Successful synthesis and failed synthesis correspond to 1 and 0 respectively. Among them, the molecular sieve data that is successfully synthesized in the sample data has a data label of 1, and the molecular sieve data label that is unsuccessful is set to 0. Among them, one-hot coding is also called one-bit effective coding, which mainly uses an N-bit state register to encode N states. Each state has an independent register bit, and only one bit is valid at any time.
[0049] Step b: using the molecular sieve synthesis experimental operating condition data as input data and the molecular sieve type data as output data to form a sample;
[0050] Step c: Perform a correlation test on the input data of the sample (i.e., calculate the correlation of the characteristics of the sample, for example, by calculating the KMO and Bartlett values to determine the correlation between the characteristics). If the calculated KMO value is greater than 0.5 and the Bartlett's sphericity test is at a significance level of 0.001, the two indicators together indicate that there is a certain correlation between the influencing factor data indicators, and thus the factor decomposition machine prediction model is used; if the KMO value and Bartlett's sphericity obtained from the correlation test do not meet the above requirements, the relevant data can be eliminated.
[0051] Step d: For the remaining sample data after removing the data that does not meet the requirements, the normalization process is performed according to the feature dimension. First, the mean μ and variance σ of all sample data under the feature are calculated, and then normalization is achieved. The normalized expression is:
[0052]
[0053] Among them, x i is the i-th experimental sample under this feature dimension, y i are normalized experimental data.
[0054] Step e: divide the normalized data set into a training set and a validation set. For example, 80% of the sample data can be divided into a training set, and 20% of the sample data can be divided into a validation set, that is, the number of samples in the training set is 4100, and the number of samples in the validation set is 1025. Of course, the present invention is not limited to this, and the training set and the validation set can be divided according to actual needs.
[0055] Step 208: Based on the data set, construct and train a molecular sieve type prediction model based on the factorization machine.
[0056] Specifically, a molecular sieve type prediction model based on the factorization machine can be constructed and trained based on the training set so that the molecular sieve type prediction model based on the factorization machine predicts the type of molecular sieve synthesized under the input operating conditions. The molecular sieve type prediction model based on the factorization machine is a second-order factorization machine.
[0057] Specifically, it may include the following steps:
[0058] Step a: Establish a molecular sieve type prediction model based on the factorization machine, and use the stochastic gradient descent method to optimize its network loss Loss, where Loss is the cross-entropy loss, and the cross-entropy loss is a commonly used loss function in the field. For the number 2 of molecular sieve types, the expression of the cross-entropy loss Loss is:
[0059]
[0060] where n represents the number of samples, y′ i represents the label value (0 or 1) of sample i, and p i represents the probability that sample i is predicted as 1;
[0061] For the molecular sieve type prediction model based on the factorization machine, the dimension of the hidden layer vector is uniformly set to 128, where the input data dimension in_dim is 68 and the output data dimension out_dim is 1.
[0062] Step b: Train the factorization machine model, that is, the molecular sieve type prediction model. When the factorization machine model is trained with the training set to reach the preset number of times, perform an error test on the validation set. When the error of the validation set rises, stop training and save the factorization machine model. In this embodiment, the preset number of times is 500 generations. Here, the model is first trained without using the validation set for verification. When the set number of times for the first-stage training is reached, the model starts to be verified with the validation set while training. If the loss of the validation set has been decreasing and the error has been decreasing, it means that the model is still learning and no overfitting has occurred. Once the error of the validation set starts to increase, it means that the model starts to overfit and the model iteration needs to be stopped immediately.
[0063] Figure 4 Schematically shows a flowchart of a method for predicting the type of molecular sieve based on a factorization machine according to an embodiment of the present application. As Figure 4 shown, the following uses the historical data modeling of high-throughput molecular sieve storage to illustrate a method for predicting the type of molecular sieve based on a factorization machine proposed by the present invention. The method includes the following steps:
[0064] Step S1: Obtain historically stored experimental operation condition data and molecular sieve type data from the high-throughput molecular sieve database, and ensure the one-to-one correspondence between each experimental batch. The molecular sieve synthesis experimental data set contains 5946 sample data;
[0065] Step S2: preprocessing the molecular sieve synthesis experimental data, removing samples with missing values and outliers, and reducing the sample size to 5174;
[0066] Step S3: In the preprocessed data set, multiple experimental sample data of the same operating condition are sorted out to obtain one-to-one corresponding operating condition conditions and molecular sieve type data, and the sample capacity is reduced to 5125.
[0067] Step S4: construct structured data for model training. First, one-hot encoding is used to encode the success or failure of the synthesis of molecular sieve types into a 0-1 form as a label for model output; the molecular sieve synthesis experimental operating condition data is used as input data, the input data dimension in_dim is 68, the input data is correlated, the KMO value and Bartlett sphericity are evaluated, and sample data that do not meet the correlation standard are eliminated; for the remaining sample data after eliminating the sample data that do not meet the correlation standard, the molecular sieve type data is used as output data, the output data dimension out_dim is 1, and 80% of the samples in the data set are divided into a training set, and 20% are divided into a validation set, that is, the number of samples in the training set is 4100, and the number of samples in the validation set is 1025;
[0068] Step S5: construct and train the factor decomposition machine model. In the network structure of the factor decomposition machine model, the hidden vector dimension is uniformly set to 128, where the input data dimension in_dim is 68, and the output data dimension out_dim is 1. When training the factor decomposition machine model, the preset number of generations is 500. Figure 3 As mentioned above, the change of the loss curve during the training of the factorization machine model is shown. The factorization machine model tends to converge after 150 generations.
[0069] Step S6: using the trained molecular sieve type prediction model based on the factor decomposition machine to predict the molecular sieve experimental data, and predicting the corresponding product molecular sieve type from the operating conditions.
[0070] Specifically, the molecular sieve experimental operating condition data is input into the molecular sieve type prediction model saved after training to predict the molecular sieve type data; for the output molecular sieve type data, that is, one-hot encoded data, the molecular sieve type set to 1 in the molecular sieve type data is used as the final predicted molecular sieve type.
[0071] The present invention utilizes the experimentally obtained data of zeolite synthesis stored in history to establish a prediction model from operating conditions to the success or failure of zeolite synthesis, predict the zeolite synthesis situation generated by experiments, provide guidance for actual zeolite synthesis experiments, and adjust the experimental operating conditions targeted, thereby reducing the R & D experimental costs.
[0072] It should be understood that although Figure 1-2 the steps in the flowchart of Figure 1-2 are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover,
[0073] An embodiment of the present application provides a storage medium, on which a program is stored, and when the program is executed by a processor, the above-mentioned zeolite type prediction method is implemented.
[0074] In one embodiment, a device for predicting the zeolite type is provided. The device can be a terminal or a server, and its internal structure diagram can be as Figure 5 shown. The device includes a processor A01, a network interface A02, a display screen A04, an input device A05, and a memory (not shown in the figure) connected through a system bus. Among them, the processor A01 of the device is used to provide computing and control capabilities. The memory of the device includes an internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A06. The network interface A02 of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor A01, a method for predicting the zeolite type is implemented. The display screen A04 of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device A05 of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.
[0075] Those skilled in the art can understand that Figure 5The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.
[0076] Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0077] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.
[0078] These computer program instructions can 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 generate a manufactured article including an instruction device, and the instruction device implements in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.
[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.
[0080] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.
[0081] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash RAM. The memory is an example of computer-readable media.
[0082] Computer-readable media includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.
[0083] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.
[0084] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.
Claims
1. A method for predicting the type of molecular sieve, characterized in that, the method comprises the following steps: obtaining the operating condition data of the molecular sieve synthesis experiment; and inputting the operating condition data of the molecular sieve synthesis experiment into a molecular sieve type prediction model based on a factorization machine to obtain product molecular sieve type data corresponding to the operating condition data of the molecular sieve synthesis experiment.
2. The method according to claim 1, characterized in that, the molecular sieve type prediction model based on a factorization machine is constructed by the following operations: obtaining molecular sieve synthesis experiment data and constructing a molecular sieve synthesis experiment data set, the molecular sieve synthesis experiment data including the operating condition data of the molecular sieve synthesis experiment and the corresponding product molecular sieve type data; based on this data set, obtaining unique operating conditions and experimental data on whether the corresponding molecular sieve is successfully synthesized; based on this experimental data, performing factor analysis on the influencing factor data, constructing structured data for model training and establishing a data set; and based on the data set, constructing and training a molecular sieve type prediction model based on a factorization machine.
3. The method according to claim 2, characterized in that, before obtaining unique operating conditions and experimental data on whether the corresponding molecular sieve is successfully synthesized, it further includes: preprocessing the molecular sieve synthesis experiment data in the data set, the preprocessing including missing value processing and outlier processing, wherein, the missing value processing includes: pairing the molecular sieve experiment operating condition data and the product molecular sieve type data one by one, and removing the sample data missing one of them; wherein, the outlier processing includes: discarding the outlier data samples, and the outlier data samples include the sample data with the temperature or pressure deviating from the set value due to the abnormality of the reaction instrument.
4. The method according to claim 2, characterized in that, obtaining unique operating conditions and experimental data on whether the corresponding molecular sieve is successfully synthesized includes: traversing the data set, comparing whether the operating conditions of different sample data are the same, if so, clustering the sample data into one category; traversing all categories containing multiple samples, judging whether the molecular sieve types in the same category are the same, if so, taking this type as the molecular sieve type corresponding to the sample data of this category, if not, discarding the sample data of this category.
5. The method according to claim 2, characterized in that, based on this experimental data, performing factor analysis on the influencing factor data, constructing structured data for model training and establishing a data set includes: using one-hot encoding to encode whether the synthesis of the molecular sieve type in the experimental data into the 0-1 form as the label output by the model; taking the molecular sieve synthesis experiment operating condition data in the experimental data as input data, calculating the correlation of the input data, and evaluating the KMO value and Bartlett sphericity; and eliminating the experimental data whose KMO value and Bartlett sphericity do not meet the preset criteria, and taking the remaining data as the structured data for model training, taking a part of the structured data as the training set, and taking another part of the structured data as the validation set.
6. The method according to claim 2, It is characterized in that Among them, Constructing and training a molecular sieve type prediction model based on a factorization machine includes: Establishing a molecular sieve type prediction model based on a factorization machine, and using the stochastic gradient descent method to optimize the network loss Loss of the molecular sieve type prediction model; and Training the molecular sieve type prediction model. When the training set trains the molecular sieve type prediction model to reach a preset number of times, perform an error test on the validation set. When the error of the validation set rises, stop training and save the molecular sieve type prediction model.
7. The method according to claim 6, It is characterized in that The hidden vector dimension of the molecular sieve type prediction model is uniformly set to 128.
8. The method according to claim 6, It is characterized in that The network loss Loss is a cross-entropy loss.
9. The method according to any one of claims 1-8, Among them, The operating condition data of the molecular sieve synthesis experiment includes one or more of the following: the types and contents of raw materials for molecular sieve synthesis, reaction temperature, reaction time, and stirring data; The product molecular sieve type data includes one or more of the following: ZSM-5, SAPO, MOR, and Beta types.
10. The method according to any one of claims 1-8, Among them, The molecular sieve type prediction model is as follows: where x i and x j represent the i-th and j-th features of each sample, w i is the learning weight of the i-th feature, w 0 is the initialized weight, n represents that each sample has n features, and these n features constitute the operation condition data of the molecular sieve synthesis experiment, v i is the hidden vector of the i-th dimensional feature, <,> represents the dot product of vectors, the length of the hidden vector is k (k << n), and it includes k factors describing the features.
11. An apparatus for predicting the type of molecular sieve, the apparatus Comprises: A memory; And A processor configured to execute the molecular sieve type prediction method according to any one of claims 1 to 10.
12. A machine-readable storage medium having instructions stored thereon, It is characterized in that When the instructions are executed by a processor, the processor is configured to execute the molecular sieve type prediction method according to any one of claims 1 to 10.