Method for optimizing production process of surfactant

Abstract

The invention provides a production process optimization method of a surfactant, and relates to the technical field of surfactants, and the production process optimization method comprises the following steps: collecting historical production data of the surfactant; constructing a production process prediction model; performing performance prediction on the initial process parameters, and comparing the initial process parameters with a target performance value; when the predicted performance difference value is larger than a preset performance difference threshold value, technological parameter optimization is executed; producing a surfactant sample by adopting the optimized process parameters, obtaining an actual performance value of the sample, and comparing the actual performance value with a target performance value; when the sample performance difference value is smaller than or equal to a preset performance difference threshold value, the optimized process parameters serve as execution process parameters for machining. By means of the method and device, the technical problem that in the prior art, due to the fact that the influence of technological parameter changes on the product performance is large, the production efficiency of the surfactant is low can be solved, the prediction model is constructed through historical data, the technological parameters are optimized, and the production efficiency and the product quality of the surfactant are improved.

Description

Technical Field

[0001] The present application relates to the technical field of surfactants, and in particular to a method for optimizing the production process of surfactants. Background Art

[0002] Surfactant is a chemical substance with both hydrophilic and hydrophobic groups in its molecule. It can reduce the surface tension or interfacial tension of liquids and is widely used in detergents, emulsifiers, lubricants and other fields. In the production process of surfactants, there are usually multiple process parameters that need to be precisely controlled, such as reaction temperature, reaction time, stirring speed, raw material ratio, reaction pressure, etc., which often have certain fluctuations and have a direct impact on the final performance of the surfactant. In the production process of surfactants, even the same batch of raw materials and roughly the same process conditions may lead to large differences in the performance of the final product due to slight changes in process parameters. Performance inconsistency directly affects the effect of surfactants in various applications, thereby affecting the overall efficiency of the production process.

[0003] At present, process parameters are usually set based on historical experience before production begins, and real-time adjustments are rarely made during the production process. Once a change or unstable factor occurs in a certain link, it is difficult to detect and correct it in time, resulting in fluctuations in product performance. Although some high-end equipment and automated production lines can achieve automatic control of certain process parameters, the lack of accurate models to predict and adjust the parameters of each link limits production efficiency and product consistency.

[0004] In summary, the prior art has a technical problem that the instability of process parameters leads to inconsistent performance of surfactant products, which further affects production efficiency. Summary of the invention

[0005] The purpose of this application is to provide a method for optimizing the production process of surfactants, so as to solve the technical problem in the prior art that the instability of process parameters leads to inconsistent performance of surfactant products, further affecting production efficiency.

[0006] In view of the above problems, the present application provides a method for optimizing the production process of a surfactant, wherein the method for optimizing the production process of a surfactant comprises: collecting historical production data of the surfactant, wherein the historical production data comprises process parameter information and performance parameter information; constructing a production process prediction model based on the process parameter information and the performance parameter information, wherein the production process prediction model is used to predict the performance value of the surfactant under the process parameters; performing performance prediction on the initial process parameters through the production process prediction model to obtain a predicted performance value, and comparing the predicted performance value with a target performance value to obtain a predicted performance difference; when the predicted performance difference is greater than a preset performance difference threshold, performing process parameter optimization with the goal of reducing the predicted performance difference to obtain optimized process parameters; producing a surfactant sample using the optimized process parameters, and obtaining the actual performance value of the sample, and comparing the actual performance value of the sample with the target performance value to obtain the sample performance difference; when the sample performance difference is less than or equal to the preset performance difference threshold, using the optimized process parameters as the execution process parameters to process and produce the surfactant.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By collecting historical production data of surfactants, the historical production data includes process parameter information and performance parameter information; based on the process parameter information and the performance parameter information, a production process prediction model is constructed, wherein the production process prediction model is used to predict the performance value of the surfactant under the process parameters; the initial process parameters are predicted by the production process prediction model to obtain the predicted performance value, and the predicted performance value is compared with the target performance value to obtain the predicted performance difference; when the predicted performance difference is greater than the preset performance difference threshold, the process parameter optimization is performed with the goal of reducing the predicted performance difference to obtain the optimized process parameters; the surfactant samples are produced using the optimized process parameters, and the actual performance value of the samples is obtained, and the actual performance value of the samples is compared with the target performance value to obtain the sample performance difference; when the sample performance difference is less than or equal to the preset performance difference threshold, the optimized process parameters are used as the execution process parameters to process and produce the surfactant. That is to say, the production process prediction model is constructed through historical data, and the process parameters are optimized according to the performance prediction results, thereby improving the production efficiency and product quality of the surfactant.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0010] Figure 1 The present invention is a schematic diagram of a process flow of a surfactant production process optimization method.

[0011] Figure 2 This is a schematic diagram of a process for obtaining optimized process parameters in a method for optimizing the production process of a surfactant in the present application. DETAILED DESCRIPTION

[0012] This application provides a surfactant production process optimization method to solve the technical problem in the prior art that the instability of process parameters leads to inconsistent surfactant product performance, which further affects production efficiency. A production process prediction model is constructed through historical data, and process parameters are optimized according to performance prediction results, thereby improving the production efficiency and product quality of surfactants.

[0013] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0014] For examples, please see the attached Figure 1 The present application provides a method for optimizing the production process of a surfactant, wherein the method for optimizing the production process of a surfactant specifically comprises the following steps: Step 1: Collect historical production data of surfactants, wherein the historical production data includes process parameter information and performance parameter information.

[0015] Specifically, data on the production of surfactants over a period of time in the past, including process parameters and performance parameters, are collected from the production database. The production database refers to the process parameters recorded in the production process from the laboratory, pilot plant or large-scale production line through the monitoring equipment or experimental record sheet of the production line, and the collected data is saved in the production database for subsequent query and analysis. In other words, by identifying the key points of process parameters and performance parameters that need to be monitored on the production line. Temperature sensors, flow meters, analytical instruments, etc. are installed at the determined collection points for real-time monitoring and recording of data. The data provided by sensors and monitoring equipment are automatically collected, stored in the database, and the data is cleaned, deduplicated and structured. The process parameters and performance parameters are established in a corresponding relationship and entered into the database together to form a production database. Process parameter information refers to indicators related to production conditions, such as temperature, pressure, raw material ratio, reaction time, etc. Performance parameter information refers to indicators related to product quality and function, such as surface tension, foaming ability, biodegradability, product purity, activity, etc. Through automated collection and sorting, the errors of manual recording are reduced, and the integrity and consistency of the collected data are ensured.

[0016] Step 2: Based on the process parameter information and the performance parameter information, a production process prediction model is constructed, wherein the production process prediction model is used to predict the performance values ​​of the surfactant under the process parameters.

[0017] Specifically, based on the process parameter information and performance parameter information, a sample process parameter set and a sample performance value set are constructed. The sample process parameter set and the sample performance value set are divided into two parts according to a preset ratio, namely a training set and a validation set. The preset ratio refers to the division ratio set according to actual needs, usually 80% training set and 20% validation set. The training set is used to train the model to help the model learn the relationship between process parameters and performance values. The validation set is used to test the generalization ability of the model to prevent the model from overfitting.

[0018] Using the training set, multiple (N) different types of basic prediction models are trained, which can be models of different algorithms, such as linear regression, support vector machine, neural network, etc., to obtain the first basic prediction model to the Nth basic prediction model. The validation set is used to verify the first basic prediction model to the Nth basic prediction model to obtain the accuracy of the N basic models. When the accuracy of all basic models meets the preset accuracy constraints, the weight of their integration is determined according to the accuracy of each model. The outputs of the first basic prediction model to the Nth basic prediction model are weighted and combined according to the integration weight to construct an output integration layer, and the output ends of the first basic prediction model to the Nth basic prediction model are connected to the input ends of the output integration layer to form a more powerful production process prediction model. The model learns and outputs prediction values ​​based on historical data, which are used to monitor and adjust process conditions in real time during the production process. The production process prediction model is used to predict the specific performance of surfactants according to different process parameters. By constructing a production process prediction model, the product performance under different process parameters is predicted, thereby optimizing the production process and improving product quality and production efficiency.

[0019] Step three: Use the production process prediction model to perform performance prediction on the initial process parameters to obtain a predicted performance value, and compare the predicted performance value with the target performance value to obtain a predicted performance difference.

[0020] Specifically, during the production process, the initial process parameters are set according to the process design or production plan. The initial process parameters refer to the process parameter values ​​initially set during the production process, such as temperature, time, concentration, etc., which will be used as the input of the model for prediction. The initial process parameters are input into the constructed production process prediction model for performance prediction to obtain predicted performance values, such as predicted activity, stability, etc. The model predicts performance values ​​based on the input parameters through existing historical data and training process. The target performance value is usually set according to product design specifications, quality standards or customer needs, and is the performance standard expected to be achieved during the production process.

[0021] Compare the predicted performance value with the target performance value to obtain the predicted performance difference, that is, the difference between the predicted performance value and the target performance value, which is used to measure the deviation between the predicted result and the target. Usually, the predicted performance difference is obtained by subtracting the target performance value from the predicted performance value. This difference reflects the accuracy of the prediction. By comparing the predicted performance difference with the preset error threshold, it is determined whether the production process needs to be adjusted. By comparing the predicted performance with the target performance, it is possible to promptly find out whether the process settings deviate from the target, which helps to quickly adjust the production parameters and ensure product quality.

[0022] Step 4: When the predicted performance difference is greater than a preset performance difference threshold, process parameter optimization is performed with the goal of reducing the predicted performance difference to obtain optimized process parameters.

[0023] Specifically, the predicted performance difference refers to the difference between the predicted performance value and the target performance value obtained when the performance of the initial process parameters is predicted through the production process prediction model. This difference is an important indicator to measure the accuracy of the prediction results, and is usually used to determine whether the current process parameter settings can meet the quality requirements of the product. The preset performance difference threshold is a set value used to determine whether the predicted performance difference is small enough. If the predicted performance difference is greater than the preset threshold, it indicates that the current process parameter settings cannot meet the requirements and need to be optimized. When the predicted performance difference is greater than the preset performance difference threshold, it means that the current process parameters have not yet met the predetermined performance requirements and process parameter optimization is required. Process parameter optimization is an iterative optimization process, the purpose of which is to reduce the predicted performance difference by adjusting the process parameters and ultimately obtain the optimized process parameters.

[0024] Based on the initial process parameters, a wider parameter space is defined, which contains possible combinations of process parameters. In the determined process parameter optimization space, multiple candidate process parameter combinations are generated. These combinations can be randomly generated or obtained through optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.). Random generation involves randomly selecting parameter values ​​in the process parameter optimization space to generate candidate process parameter combinations. The value of each parameter is randomly selected from the possible range defined by it, which is simple and easy to implement, but requires a large number of random combinations to cover a sufficient part of the parameter space to ensure that the optimal or better parameter combination is found. Using optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.) to generate candidate process parameter combinations, usually based on the principles of natural selection and evolution, can search more intelligently in the parameter space to find more potential parameter combinations. For example, genetic algorithms simulate the process of natural selection, select and combine the best performing parameter combinations, and generate new candidate combinations. The particle swarm optimization algorithm simulates the behavior of bird flocks or fish schools, and finds the optimal solution through the interaction and sharing of information among individuals in the group. It can find a better parameter combination in a fewer number of iterations, improving the efficiency and accuracy of optimization.

[0025] Use the production process prediction model to predict the performance of each candidate process parameter combination to obtain a predicted performance value, and perform weighted calculation on the predicted performance value and the actual performance mean corresponding to the candidate process parameter combination to obtain multiple candidate predicted performance values. Sort the multiple candidate predicted performance values ​​according to their closeness to the target performance value, select the optimal predicted performance value, and continuously generate candidate process parameters and predict performance in the subspace of the process parameter optimization space according to the corresponding candidate process parameters to obtain multiple local candidate predicted performance values. Among the multiple candidate predicted performance values ​​and multiple local candidate predicted performance values, determine a set of optimized process parameters and use them as the optimized process parameters, which represent the parameter combination found in the optimization process that is most likely to meet the target performance requirements. Through process parameter optimization, the production process can be optimized, the product performance instability caused by process fluctuations can be reduced, and the quality of each batch of products can be ensured to be close to or meet the target requirements.

[0026] Step 5: Produce a surfactant sample using the optimized process parameters, and obtain an actual performance value of the sample, compare the actual performance value of the sample with the target performance value, and obtain a sample performance difference.

[0027] Specifically, production is carried out according to the optimized process parameters, including adjusting reaction conditions such as temperature, time, stirring speed, etc., to produce surfactant samples. Surfactant samples refer to surfactant products produced according to the optimized process parameters. After optimizing the process parameters, the samples produced are used to verify whether the actual production effect is consistent with the target performance. The target performance value is an ideal performance indicator preset in the production process, which is usually determined according to market demand, product standards or user requirements. The target performance value of the surfactant may include surface tension, foam performance, solubility, etc. The actual performance value of the sample is compared with the pre-set target performance value, and the difference between the actual performance value and the target performance value is calculated, that is, the actual performance value is subtracted from the target performance value. The smaller the sample performance difference, the better the optimization effect, and the product performance is close to the target requirements. By comparing the actual production sample test with the target performance value, the effect of optimizing the process parameters can be verified. If the difference is small, it means that the optimized process parameters are effective, otherwise they need to be adjusted. By collecting and comparing the sample performance data in real time, it is possible to timely discover the deficiencies in the optimization process, make adjustments quickly, and reduce the number of unqualified products in production.

[0028] Step six: When the sample performance difference is less than or equal to the preset performance difference threshold, the optimized process parameters are used as execution process parameters to process and produce surfactants.

[0029] Specifically, the preset performance difference threshold refers to a maximum tolerance difference set according to production requirements, quality control standards or market demand. If the sample performance difference is less than or equal to the threshold, it means that the product quality meets the requirements and can be put into production. When the sample performance difference meets the requirements, the process parameters are regarded as the final execution process parameters and will be used for large-scale production to ensure that the produced surfactant meets the target performance value. According to the verified qualified execution process parameters, the production line is started for large-scale production of surfactants. During the production process, the reaction conditions need to be precisely controlled and the operation is carried out according to the optimized process parameters to ensure that each batch of products meets the expected performance standards. By directly carrying out large-scale production after confirming the optimized process parameters, unnecessary adjustments and re-optimization can be avoided, the generation of defective products can be reduced, thereby reducing production costs and material waste. After confirming that the sample performance meets the requirements, the optimized process parameters are directly applied for production, avoiding the waste of time in repeatedly adjusting the process, thereby improving the overall production efficiency.

[0030] Furthermore, the present application also includes the following steps: When the sample performance difference is greater than the preset performance difference threshold, the optimized process parameters are taken as a starting point and optimized in combination with the production process prediction model until the execution process parameters are generated.

[0031] Specifically, if the sample performance difference is greater than the preset threshold, it means that the current process parameters fail to meet the quality requirements and need further optimization. The optimized process parameters are adjusted in the process parameter optimization space to generate new candidate process parameters. After each adjustment, the model is used to predict the new sample performance, the performance difference is calculated, and whether to continue optimization is determined based on whether the difference meets the standard. This process requires multiple iterations until the generated process parameters can make the sample performance difference less than or equal to the preset performance difference threshold. The specific optimization process is similar to the process parameter optimization, which is described in detail in the refinement of the subsequent steps, and will not be repeated here. When the performance difference of the sample drops below the preset threshold after multiple optimizations and predictions, the generated process parameters are the execution process parameters. These final process parameters can be applied to actual production lines. The execution process parameters are the process parameters that can be put into large-scale production after multiple optimizations, ensuring that the produced surfactants can reach the target performance values. Through the collaboration of the optimization algorithm and the prediction model, it is ensured that the optimized process parameters are stable and meet the expected performance requirements. By continuously optimizing process parameters until the execution process parameters are generated, it is ensured that the products produced meet the predetermined performance requirements, reduce the cost of experiments and trial and error, improve the flexibility and adaptability of the production process, and allow producers to adjust process parameters according to different goals and constraints to optimize product performance and production efficiency. By combining the production process prediction model for optimization, the best or better process parameter combination can be quickly found, thereby optimizing the production process, improving product quality, and meeting market demand.

[0032] Further, step 2 of this application includes: Acquire a set of sample process parameters and a set of sample performance values, wherein the set of sample process parameters is constructed by the process parameter information, and the set of sample performance values ​​is constructed by the performance parameter information; divide the set of sample process parameters and the set of sample performance values ​​into a training set and a validation set according to a preset ratio; use the training set to train N different types of basic prediction models to obtain a first basic prediction model to an Nth basic prediction model, wherein N is an integer greater than 2; use the validation set to verify the first basic prediction model to the Nth basic prediction model to obtain the accuracy rates of N basic models; when the accuracy rates of the N basic models all meet the preset accuracy constraints, integrate the first basic prediction model to the Nth basic prediction model to obtain the production process prediction model.

[0033] Specifically, the process parameter information is constructed as a sample process parameter set, and the performance parameter information is constructed as a sample performance value set. The sample process parameter set includes a series of process parameter combinations, and the sample performance value set includes performance parameter values ​​corresponding to the process parameters. The constructed sample process parameter set and sample performance value set are divided into two parts according to a preset ratio: a training set and a validation set. The training set is used to build and train the production process prediction model, while the validation set is used to evaluate the accuracy and generalization ability of the model. The preset ratio refers to a ratio set in advance to determine the number of samples contained in each set, which is usually determined based on the total amount of data and the needs of model training, usually 80% training set and 20% validation set.

[0034] Use the previously divided training set to train multiple different types of basic prediction models. Each model is built based on different algorithms or architectures in order to obtain different prediction perspectives on surfactant performance, which can be decision trees, random forests, support vector machines, neural networks, etc. Each basic prediction model is a single type of machine learning model, which is trained on the training set and outputs a specific prediction result. For example, the linear regression model assumes that there is a linear relationship between the feature and the target variable; the support vector machine (SVM) predicts the target value by finding a hyperplane that maximizes the decision boundary; random forest (RF): based on the integration of multiple decision trees, it improves the accuracy and robustness of the prediction. Selecting multiple different types of models (N>2) is to improve the stability and adaptability of the model prediction. The specific training process is to extract feature variables (process parameter set) and target variables (performance parameter set) from the training set, preprocess the data, select N different types of machine learning algorithms, and fit each model using feature variables and target variables. After the training is completed, the first basic prediction model to the Nth basic prediction model are obtained.

[0035] The validation set is used to validate the first to Nth basic prediction models, and the process parameters of the validation set are input into the first to Nth basic prediction models. The prediction performance value of the model is compared with the performance parameter value in the validation set, and the accuracy of the model is evaluated to obtain the accuracy of the N basic models. Accuracy is an indicator that measures the degree of closeness between the model prediction result and the true value, which can be calculated by the coefficient of determination, mean absolute error, etc. The coefficient of determination (R²) indicates the proportion of variability explained by the model, and its value ranges from 0 to 1. The closer it is to 1, the better the model fit; the mean absolute error (MAE) indicates the size of the average difference between the predicted value and the actual value. The lower its value, the more accurate the model prediction.

[0036] Compare the accuracy of N basic models and select models that meet the performance requirements. According to the accuracy of the basic model, configure the corresponding integration weight. The model with higher accuracy will get a higher weight. The output of each basic model is weighted and combined according to its corresponding weight to obtain the output integration layer. The output of the first basic prediction model to the Nth basic prediction model is connected to the input of the output integration layer to obtain a complete production process prediction model. Integration refers to improving the overall prediction performance by merging the predictions of multiple models.

[0037] When all N basic models fail to meet the preset accuracy constraints, measures need to be taken to improve the model or data to improve the accuracy and ensure that the basic model has sufficient predictive ability. The measures taken include improvements at the data level, model level, and integration strategy. Improvements at the data level can increase training sample data, mine unused production data from historical records, or further collect samples through experiments to more comprehensively learn the relationship between process parameters and performance. Some process parameters may have a weak relationship with performance, making it difficult to accurately fit the model. Important features can be screened or new combined features can be constructed to improve performance. Improvements at the model level Since hyperparameters directly affect the learning ability of the model, optimizing hyperparameters can improve model performance. Or some models may not be suitable for the characteristics of the data, and introducing models with better performance can improve the overall accuracy. Improvements in integration strategies can further improve performance through more complex integration methods, because simple weighted averages may not fully utilize the information of the basic models. Through the integration of basic models, the advantages of the model can be maximized and the prediction accuracy can be improved. Multi-model integration effectively reduces the impact of outliers or errors and performs more robustly.

[0038] Furthermore, the present application also includes the following steps: Based on the accuracy rates of the N basic models, N integration weights are configured, wherein the integration weights are positively correlated with the accuracy rates of the basic models; using the N integration weights, an output integration layer is constructed, wherein the output integration layer is used to perform weighted combination of the outputs of the first basic prediction model to the Nth basic prediction model; connecting the output ends of the first basic prediction model to the Nth basic prediction model to the input ends of the output integration layer to construct the production process prediction model.

[0039] Specifically, the preset accuracy constraint refers to the set minimum accuracy threshold. Only when the prediction performance (such as R² or MAE) of each basic model meets this threshold, the model integration stage is allowed to enter. If the accuracy of all models meets the preset accuracy constraint, that is, the set performance standard is reached, then the integration weight will be configured based on the accuracy of these models. The integration weight is the importance weight assigned to each basic model, and its value is positively correlated with the accuracy of the model, that is, the model with higher accuracy will get a higher weight. For example, if the accuracy of the linear regression model is 85%, the decision tree model is 88%, and the neural network model is 90%, then the weight of the neural network model is greater than the other two models. Assume that the weight calculation formula is: weight = (model accuracy / sum of all model accuracies). According to the above accuracy, the weight of the linear regression model is 0.32, the weight of the decision tree model is 0.33, and the weight of the neural network model is 0.34.

[0040] The output of each basic model is weighted and combined according to its corresponding weight to obtain a comprehensive prediction result and construct an output integration layer. The output end of each basic model is connected to the input end of the output integration layer to form a complete production process prediction model. When new process parameters are input into these basic models, each generates a performance prediction value, which is input into the output integration layer. The prediction results of each basic model will be considered by the integration layer, and weighted combination will be performed according to the corresponding weights to finally output a comprehensive prediction result. By integrating the prediction results of multiple basic models, the advantages of different models can be fully utilized to improve the overall prediction ability. Weighted averaging reduces the impact of individual model errors on the overall prediction and improves the adaptability of the model to new samples.

[0041] Furthermore, the present application also includes the following steps: According to a preset time interval, multiple cycle predicted performance values ​​and multiple cycle actual performance values ​​within the preset time interval are collected; the degree of deviation between the multiple cycle predicted performance values ​​and the multiple cycle actual performance values ​​is calculated; when the degree of deviation is greater than a prediction deviation threshold, the update accuracy of the first basic prediction model to the Nth basic prediction model is calculated based on the multiple cycle actual performance values; the update accuracy is used to reconfigure the integration weights of the output integration layer, and update the production process prediction model.

[0042] Specifically, the preset time interval refers to the time interval for collecting data at a fixed time period during the production process. A time period is preset to monitor and adjust the performance of the prediction model. According to the preset time interval (such as daily, weekly or monthly), collect multiple period prediction performance values ​​obtained by the prediction model within the time interval and multiple period actual performance values ​​obtained in actual production. Calculate the degree of deviation between the multiple period prediction performance values ​​and the multiple period actual performance values, using indicators such as mean square error (MSE) and mean absolute error (MAE). Mean absolute error (MAE) refers to the average of the absolute values ​​of the deviations between all period prediction values ​​and actual values; mean square error (MSE) refers to the average of the squares of the deviations between all period prediction values ​​and actual values.

[0043] The prediction deviation threshold refers to the maximum deviation allowed. If the deviation exceeds the threshold, the model needs to be adjusted. If the deviation is greater than the prediction deviation threshold, it means that the prediction performance of the model is poor and needs to be updated. Recalculate the accuracy of the first basic prediction model to the Nth basic prediction model based on the actual performance values ​​collected for multiple cycles to obtain the updated accuracy. Reconfigure the integration weights of the output integration layer according to the updated accuracy to ensure that the integrated model can better reflect the current production status and improve the accuracy and reliability of the prediction. Reconfiguring the integration weights involves adjusting the contribution of each basic model in the integrated model so that the model with higher accuracy occupies a larger weight in the final prediction. Update the production process prediction model through the reconfigured integration weights to improve the prediction accuracy. By collecting data and calculating the deviation degree, the performance degradation of the production process prediction model can be discovered in time. When the deviation degree exceeds the preset threshold, the prediction model is dynamically adjusted and optimized by calculating the updated accuracy and reconfiguring the integration weights to ensure that it always has a high prediction accuracy, which helps to continuously improve the production process, reduce uncertainty in the production process, and improve product quality and production efficiency.

[0044] Further, as attached Figure 2 As shown, step 4 of this application includes: Based on the initial process parameters, a process parameter optimization space is configured, and a plurality of candidate process parameters are generated in the process parameter optimization space; the production process prediction model is used to perform performance prediction on the plurality of candidate process parameters to obtain a plurality of candidate prediction performance values; and the optimized process parameters are obtained based on the plurality of candidate prediction performance values ​​and the process parameter optimization space.

[0045] Specifically, when the predicted performance difference is greater than the preset performance difference threshold, it indicates that the predicted value deviates greatly from the target performance, and the process parameter optimization process needs to be started. The preset performance difference threshold is a pre-defined tolerance range used to determine whether the difference between the predicted performance value and the target performance value is acceptable. According to the initial process parameters, the adjustable range of each process parameter is determined, and the process parameter optimization space is configured, which contains possible process parameter combinations. The process parameter optimization space should be large enough to include potential better parameters, but small enough to maintain the efficiency of the optimization process. The process parameter optimization space is a range of possible process parameter values ​​used to explore and select better process parameters. Within the set process parameter optimization space, multiple candidate process parameter combinations are generated through uniform sampling, random sampling, or grid search for subsequent performance prediction.

[0046] For each of the multiple candidate process parameters, a parameter search is performed in the historical production data, and one candidate process parameter is arbitrarily selected as the first candidate process parameter. Taking the first candidate process parameter as an example, multiple first historical actual performance values ​​corresponding to it are obtained, and the average of the multiple first historical actual performance values ​​is calculated to obtain the first actual performance average. The first candidate process parameter is input into the production process prediction model for performance prediction to obtain the first predicted performance value. The first predicted performance value and the first actual performance average are weightedly calculated to obtain the corresponding first candidate predicted performance value. The above calculation is also performed on other candidate process parameters in the multiple candidate process parameters to obtain multiple candidate predicted performance values.

[0047] Sort multiple candidate prediction performance values ​​according to their closeness to the target performance value, with the closer ones being ranked at the front, and select the first few preferred prediction performance values, that is, the prediction performance values ​​closest to the target performance value. According to the candidate process parameters corresponding to the preferred prediction performance values, configure a local optimization space in the process parameter optimization space, that is, a smaller area selected from the process parameter optimization space, and the process parameters in this area are more finely concentrated in the area with better prediction performance values. Further optimization in this subspace can help to accurately find the optimal process parameters. Continuously iterate in the local optimization space, and each iteration will generate new candidate process parameters based on the results of the previous iteration and predict them.

[0048] By repeatedly executing the above process, new candidate process parameters are gradually generated and their performance is predicted. Each round of iteration will update the local optimization space and further optimize the process parameters until the preset number of iterations is reached. Through multiple iterative optimizations, multiple candidate prediction performance values ​​and local candidate prediction performance values ​​are evaluated, and the optimal process parameters are selected, that is, the process parameters are optimized to produce the best surfactant performance in the production process. Through the configuration of the local optimization space, the search range can be narrowed to avoid invalid searches in the entire process parameter space, thereby greatly improving the efficiency of the optimization process. Generating candidate process parameters and performing performance predictions can provide multiple groups of optimization solutions in a short period of time, quickly screen out the optimal process parameters, save experimental and debugging time, and can flexibly adapt to different production needs and conditions. It is suitable for a variety of production environments and is not limited to a specific production line or product.

[0049] Furthermore, the present application also includes the following steps: The multiple candidate predicted performance values ​​are sorted according to their proximity to the target performance value, and a preset number of preferred predicted performance values ​​are selected; based on the candidate process parameters corresponding to the preferred predicted performance values, a local optimization space is configured with the process parameter optimization space as a constraint, wherein the local optimization space is a subspace of the process parameter optimization space; within the local optimization space, candidate process parameter generation and performance prediction are repeatedly performed to obtain multiple local candidate predicted performance values; the process is iterated until a preset number of iterations is reached, and the optimized process parameters are obtained based on the multiple candidate predicted performance values ​​and the multiple local candidate predicted performance values.

[0050] Specifically, all candidate predicted performance values ​​are sorted according to their closeness to the target performance value, and the degree of closeness can usually be measured by absolute error (i.e., the difference between the target performance value and the predicted value). A preset number of predicted performance values ​​that are closest to the target performance value are selected as preferred performance values. The preset number is a pre-set number that depends on the sophistication of the optimization and the computing resources. The preferred predicted performance value refers to the predicted performance value that is closest to the target performance value. The target performance value refers to the product performance index that is expected to be achieved in the production process of the surfactant, such as surface tension, foam stability, etc. It is the reference standard for the optimization process. The goal is to make the final product performance value as close to the target performance value as possible by adjusting the process parameters.

[0051] According to the candidate process parameters corresponding to the preferred predicted performance values, a small range in the process parameter space is selected as the local optimization space. In this local optimization space, the changes in process parameters will be more refined, which will help to find the optimal process parameter combination in a smaller range. The local optimization space is a subset of the process parameter optimization space, which narrows the scope of optimization and focuses on those parameter areas that have shown good performance. In the local optimization space, multiple new candidate process parameter combinations are generated randomly or regularly, and the multiple new candidate process parameter combinations are input into the production process prediction model for performance prediction, and multiple local candidate prediction performance values ​​are obtained, which represent the product performance prediction under different process parameter combinations in the local optimization space. Perform multiple iterations in the local optimization space, and each iteration will generate new candidate process parameters and make predictions. Iteration stops until the preset maximum number of iterations is reached or converges to a certain degree of optimization.

[0052] The preset number of iterations is the maximum number of times set in advance, which determines how many iterations the optimization process will perform. The setting of the number of iterations needs to consider multiple factors, including the complexity of the optimization, the availability of computing resources, and the constraints of production time. The preset number of iterations is set to ensure that the optimization process will neither end too early, resulting in failure to find the optimal solution, nor be performed too many times, wasting computing resources and time. Each iteration generates new local candidate prediction performance values, which will be used for evaluation and selection together with the previous multiple candidate prediction performance values. Based on multiple candidate prediction performance values ​​and multiple local candidate prediction performance values, a set of optimized process parameters is determined, representing the parameter combination found in the optimization process that is most likely to meet the target performance requirements. By performing fine-grained optimization in the local optimization space, the optimal process parameters can be found more accurately, avoiding the inefficiency caused by a large-scale search, and concentrating resources on optimizing the parameter area that is most likely to achieve the target. Through local optimization, unnecessary calculations and experiments are reduced, especially when the parameter space is very large. The optimized process parameters can better control the changes in the production process and ensure that each batch of products can meet consistent high quality standards.

[0053] Furthermore, the present application also includes the following steps: Traverse multiple candidate process parameters to obtain a first candidate process parameter; perform parameter retrieval in the historical production data according to the first candidate process parameter to obtain multiple first historical actual performance values ​​matching the first candidate process parameter; calculate the mean of the multiple first historical actual performance values ​​to obtain a first actual performance mean; perform performance prediction on the first candidate process parameter based on the production process prediction model to obtain a first predicted performance value; perform weighted calculation on the first actual performance mean and the first predicted performance value to obtain a first candidate predicted performance value, and add it to the multiple candidate predicted performance values.

[0054] Specifically, traverse multiple candidate process parameters, that is, visit and consider each candidate process parameter one by one, and these combinations are pre-selected from the process parameter space, representing different production configurations. By traversing these parameter combinations, subsequent operations are performed one by one. Taking the first candidate process parameter as an example, parameter retrieval is performed in the historical production data, that is, based on the process parameters in the historical production data, all historical records matching the current first candidate process parameter are found to obtain the actual performance value corresponding thereto. The first historical actual performance value refers to the multiple actual performance data corresponding to the use of the first candidate process parameter combination in the historical production data, reflecting the actual production results. Calculate the mean of the first historical actual performance value, that is, average the multiple first historical actual performance values ​​to obtain the average performance value, reflecting the typical performance of the first candidate process parameter. The mean calculation is to sum these performance values ​​and then divide them by the number of performance values ​​to obtain a representative value, namely, the first actual performance mean.

[0055] The first candidate process parameter is input into the production process prediction model. Based on the previously learned relationship between the process parameter and the performance, the model outputs a predicted performance value, namely the first predicted performance value, which represents the product performance predicted by the model under the first candidate process parameter setting. The historical performance mean and the predicted performance value are weighted to obtain the final candidate predicted performance value. The weighting method can adjust the weight according to actual needs so that the influence of historical data or predicted data in the calculation is different. For example, assuming that the multiple first historical actual performance values ​​are 34.5 millinewton / meter, 34.6 millinewton / meter, and 34.8 millinewton / meter, the first actual performance mean is calculated to be 34.63 millinewton / meter. The first predicted performance value predicted by the production process prediction model is 34.7 millinewton / meter. Assuming that the weight of the historical performance mean is 0.6 and the weight of the predicted performance value is 0.4, the first candidate predicted performance value is calculated to be 0.6×34.63+0.4×34.7=34.654 millinewton / meter.

[0056] The candidate predicted performance value refers to the final performance value obtained by combining the actual historical performance mean and the predicted performance value for each candidate process parameter combination, which is used to select the optimal process parameters. The above calculation is performed for each candidate process parameter to obtain the corresponding candidate predicted performance value, which is added to multiple candidate predicted performance values. By combining historical data and predicted data for weighted calculation, the accuracy of the final performance value is improved by integrating past experience and the prediction results of the model. The weighted calculation combines the advantages of historical actual performance and predicted values, improves the prediction ability of unknown data, and enhances the reliability of the production process prediction model.

[0057] In summary, the method for optimizing the production process of a surfactant provided in this application has the following technical effects: By collecting historical production data of surfactants, the historical production data includes process parameter information and performance parameter information; based on the process parameter information and the performance parameter information, a production process prediction model is constructed, wherein the production process prediction model is used to predict the performance value of the surfactant under the process parameters; the initial process parameters are predicted by the production process prediction model to obtain the predicted performance value, and the predicted performance value is compared with the target performance value to obtain the predicted performance difference; when the predicted performance difference is greater than the preset performance difference threshold, the process parameter optimization is performed with the goal of reducing the predicted performance difference to obtain the optimized process parameters; the surfactant samples are produced using the optimized process parameters, and the actual performance value of the samples is obtained, and the actual performance value of the samples is compared with the target performance value to obtain the sample performance difference; when the sample performance difference is less than or equal to the preset performance difference threshold, the optimized process parameters are used as the execution process parameters to process and produce the surfactant. That is to say, the production process prediction model is constructed through historical data, and the process parameters are optimized according to the performance prediction results, thereby improving the production efficiency and product quality of the surfactant.

[0058] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0059] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for optimizing the production process of a surfactant, characterized in that: include: Collecting historical production data of surfactants, wherein the historical production data includes process parameter information and performance parameter information; Based on the process parameter information and the performance parameter information, constructing a production process prediction model, wherein the production process prediction model is used to predict the performance value of the surfactant under the process parameters; Performing performance prediction on the initial process parameters through the production process prediction model to obtain a predicted performance value, and comparing the predicted performance value with the target performance value to obtain a predicted performance difference; When the predicted performance difference is greater than a preset performance difference threshold, the process parameters are optimized to obtain the optimized process parameters with the goal of reducing the predicted performance difference; Producing a surfactant sample by using the optimized process parameters, and obtaining an actual performance value of the sample, and comparing the actual performance value of the sample with a target performance value to obtain a performance difference of the sample; When the sample performance difference is less than or equal to the preset performance difference threshold, the optimized process parameters are used as execution process parameters to process and produce the surfactant.

2. The method for optimizing the production process of a surfactant according to claim 1, characterized in that: Based on the process parameter information and the performance parameter information, a production process prediction model is constructed, including: Acquire a sample process parameter set and a sample performance value set, wherein the sample process parameter set is constructed by the process parameter information, and the sample performance value set is constructed by the performance parameter information; Dividing the sample process parameter set and the sample performance value set into a training set and a validation set according to a preset ratio; Using the training set, training N basic prediction models of different types to obtain first to Nth basic prediction models, where N is an integer greater than 2; Using the verification set to verify the first to Nth basic prediction models, and obtaining the accuracy rates of N basic models; When the accuracy rates of the N basic models all meet the preset accuracy constraints, the first basic prediction model to the Nth basic prediction model are integrated to obtain the production process prediction model.

3. The method for optimizing the production process of a surfactant according to claim 2, characterized in that: Integrating the first basic prediction model to the Nth basic prediction model to obtain the production process prediction model includes: Based on the N basic model accuracies, N integration weights are configured, wherein the integration weights are positively correlated with the basic model accuracies; Using the N integration weights, constructing an output integration layer, the output integration layer is used to perform weighted combination on the outputs of the first basic prediction model to the Nth basic prediction model; The output end of the first basic prediction model to the Nth basic prediction model is connected to the input end of the output integration layer to construct the production process prediction model.

4. The method for optimizing the production process of a surfactant according to claim 3, characterized in that: The N integration weights are used to construct an output integration layer, including: According to a preset time interval, collecting a plurality of periodic predicted performance values ​​and a plurality of periodic actual performance values ​​within the preset time interval; Calculating the degree of deviation between the predicted performance values ​​of the multiple cycles and the actual performance values ​​of the multiple cycles; When the degree of deviation is greater than a prediction deviation threshold, calculating update accuracy of the first basic prediction model to the Nth basic prediction model based on the multiple periodic actual performance values; The update accuracy is used to reconfigure the integration weight of the output integration layer and update the production process prediction model.

5. The method for optimizing the production process of a surfactant according to claim 1, characterized in that: With the goal of reducing the predicted performance difference, process parameter optimization is performed to obtain the optimized process parameters, including: Based on the initial process parameters, configuring a process parameter optimization space, and generating a plurality of candidate process parameters in the process parameter optimization space; Using the production process prediction model to perform performance prediction on the multiple candidate process parameters to obtain multiple candidate prediction performance values; The optimized process parameters are obtained based on a plurality of candidate predicted performance values ​​and a process parameter optimization space.

6. The method for optimizing the production process of a surfactant according to claim 5, characterized in that: The optimized process parameters are obtained based on a plurality of candidate predicted performance values ​​and a process parameter optimization space, including: Sort the multiple candidate prediction performance values ​​according to their proximity to the target performance value, and select a preset number of preferred prediction performance values; Based on the candidate process parameters corresponding to the preferred predicted performance values, a local optimization space is configured with the process parameter optimization space as a constraint, wherein the local optimization space is a subspace of the process parameter optimization space; Repeating the generation of candidate process parameters and performance prediction in the local optimization space to obtain a plurality of local candidate prediction performance values; The iteration is performed until a preset number of iterations is reached, and the optimized process parameters are obtained based on the multiple candidate prediction performance values ​​and the multiple local candidate prediction performance values.

7. The method for optimizing the production process of a surfactant according to claim 5, characterized in that: Also includes: Traversing multiple candidate process parameters to obtain a first candidate process parameter; Performing parameter retrieval in the historical production data according to the first candidate process parameter to obtain a plurality of first historical actual performance values ​​matching the first candidate process parameter; Calculate the average of multiple first historical actual performance values ​​to obtain the first actual performance average; Performing performance prediction on the first candidate process parameter based on the production process prediction model to obtain a first predicted performance value; A weighted calculation is performed on the first actual performance mean and the first predicted performance value to obtain a first candidate predicted performance value, which is added to the multiple candidate predicted performance values.

8. The method for optimizing the production process of a surfactant according to claim 1, characterized in that: When the sample performance difference is greater than the preset performance difference threshold, the optimized process parameters are taken as a starting point and optimized in combination with the production process prediction model until the execution process parameters are generated.

Images