CN-121999942-A - Method and system for intelligently blending and optimizing dry materials and asphalt in carbon anode production

CN121999942ACN 121999942 ACN121999942 ACN 121999942ACN-121999942-A

Abstract

The application relates to a method and a system for intelligently blending and optimizing dry materials and asphalt in carbon anode production, which belong to the technical field of carbon prebaked anode production, and the method comprises the steps of acquiring an initial data set based on orthogonal experimental design and analyzing factors; the method comprises the steps of constructing a complex relation between a Gaussian process model representation formula and multiple physicochemical properties, utilizing a Bayesian optimization algorithm, taking expected lifting as an acquisition function under process constraint to efficiently optimize until convergence conditions are met, introducing an active learning strategy to explore a high uncertainty region and a performance boundary, deepening process cognition, carrying out performance prediction with uncertainty quantification on a new formula based on a trained model, and outputting production suggestions according to a preset decision rule. According to the application, experimental design, agent modeling, active optimization and decision support are organically combined, and the global optimal formula is accurately found with the minimum experiment times, so that the research and development cost is reduced, the product quality is improved, and a complete technical solution is provided for realizing the intellectualization of carbon anode production.

Inventors

Jiang Huixiu
WANG HUAIMIN
CHEN LIKANG
YIN CHENGWEI
LI JINHUI
ZHANG XIN
HAN MINGLU
JI GUANGDONG

Assignees

济南万瑞炭素有限责任公司

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (10)

1. The intelligent blending optimization method for the dry materials and asphalt in the carbon anode production is characterized by comprising the following steps of: Step S1, acquiring an initial data set of physicochemical properties of the carbon anode under the proportion of different dry material granularity compositions and coal tar pitch based on orthogonal experimental design, and sequencing the influence importance of all factors, wherein the physicochemical properties at least comprise bulk density, compressive strength, flexural strength and room temperature resistivity; S2, taking the initial data set as a training sample, and constructing a Gaussian process model taking dry material granularity composition and coal asphalt ratio as input and taking physical and chemical properties of a carbon anode as output; Step S3, based on a Bayesian optimization framework, taking an expected lifting EI as an acquisition function, under the guidance of the Gaussian process model, iteratively searching a formula which enables the function of the comprehensive physicochemical property of the carbon anode to be optimal under the constraint that the sum of the percentages of all the granularity components is equal to 100% and the percentages of all the components are within a preset production range until a preset convergence condition is met, wherein the convergence condition comprises that the maximum evaluation times are reached or the relative improvement quantity of the optimal observation value in the continuous k iterations is smaller than a threshold delta; step S4, based on an active learning strategy, performing a targeted experiment on a region with high prediction uncertainty in the Gaussian process model, updating the Gaussian process model to reduce global uncertainty, and exploring critical conditions of performance change; And S5, based on the trained Gaussian process model, carrying out physical and chemical property prediction and uncertainty evaluation on the new candidate formula, and outputting production decision suggestions according to a preset decision rule.
2. The method for intelligent blending optimization of dry materials and asphalt for carbon anode production according to claim 1, wherein the step S1 comprises the following steps: Step S11, selecting the mass percentage of coarse particles, the mass percentage of medium particles, the mass percentage of fine particles and the mass percentage of coal tar pitch as four experimental factors, and setting three levels for each factor; step S12, selecting based on 4 factor 3 level An orthogonal table for determining the test times as 9 times; Step S13, performing carbon anode trial production tests according to an orthogonal table arrangement, and measuring and recording the bulk density, compressive strength, flexural strength and room temperature resistivity of each formula; step S14, calculating an average value K of the physicochemical property indexes corresponding to each factor at each level; Step S15, calculating the extremely poor R of each factor on each physical and chemical performance index, and sorting the influence importance of each factor according to the magnitude of the R value; And S16, selecting a formula with optimal performance, and taking the value range of each factor as an initial search space for subsequent Bayesian optimization.
3. The method for intelligent blending optimization of dry materials and asphalt for carbon anode production according to claim 1, wherein the step S2 comprises the following steps: Step S21, carrying out normalization processing on a formula-performance data set obtained by an orthogonal experiment; step S22, setting a mean function of a Gaussian process as a zero-mean function and a covariance function as a radial basis function kernel; step S23, optimizing a Gaussian process super-parameter by maximizing an edge likelihood function based on the preprocessed data set; And step S24, for any new formula point, calculating posterior distribution based on the updated Gaussian process model, and obtaining a prediction mean value and a prediction variance.
4. The method for intelligent blending optimization of dry materials and asphalt for carbon anode production according to claim 1, wherein the step S3 comprises the following steps: Step S31, taking a preferred formula range obtained by an orthogonal experiment as a Bayesian optimized search space, and taking orthogonal experiment data as an initial observation set; step S32, updating a Gaussian process model based on the current observation set; step S33, calculating expected lifting EI values of each point in the search space based on the updated Gaussian process model; step S34, selecting the point with the maximum expected lifting value as the next formula to be tested; Step S35, judging whether the selected formula meets the process constraint that the sum of the percentages of all the granularity components is equal to 100 percent and the percentages of all the components are in a reasonable production range; step S36, carrying out an actual production test on the formula meeting the constraint, and measuring the physicochemical properties of the corresponding carbon anode; Step S37, adding the formula-performance data of the new test into an observation set, and iteratively executing the steps S32 to S37 until the convergence condition is met; step S38, outputting the global optimal formula and the corresponding Gaussian process model.
5. The method for intelligent blending optimization of dry materials and asphalt for carbon anode production according to claim 1, wherein the step S4 comprises the following steps: step S41, calculating posterior standard deviations of each point on the whole formula space based on a Gaussian process model obtained by Bayesian optimization; Step S42, identifying the area with the posterior standard deviation higher than the set threshold as a high uncertainty area; step S43, a point with the largest posterior standard deviation in a high uncertainty area or a point with a predicted value close to a performance threshold and higher uncertainty is preferentially selected for testing; Step S44, adding new test data into the training set, retraining the Gaussian process model to reduce global uncertainty and explore performance boundaries.
6. The method for intelligent blending optimization of dry materials and asphalt for carbon anode production according to claim 1, wherein the step S5 comprises the following steps: step S51, receiving new formula parameters to be evaluated input by a user; step S52, carrying out normalization processing on the new formula parameters, wherein the normalization processing is the same as that of training data; Step S53, inputting the new preprocessed formula into a trained Gaussian process model to obtain a prediction mean value and a prediction standard deviation of each physicochemical property index; Step S54, calculating the uncertainty level of the new formula prediction based on the prediction standard deviation, wherein the uncertainty level is divided into low, medium and high levels according to the percentage of the prediction standard deviation to the standard deviation of the performance index training set; step S55, calculating the confidence interval of each performance index based on the prediction mean and the prediction standard deviation; and step S56, evaluating the new formula according to the decision rule and outputting advice.
7. The method for intelligently blending and optimizing carbon anode production drier and asphalt according to any one of claims 1-6, wherein the decision rule is that if the prediction mean value of all key performance indexes meets the requirement and the prediction uncertainty is low, direct application is recommended, if the prediction mean value meets the requirement but there is medium-high uncertainty, application after verification is recommended, and if the prediction mean value of any key performance index does not meet the requirement, application is not recommended, wherein the key performance indexes comprise bulk density, compressive strength, flexural strength and room temperature resistivity.
8. A system for intelligent blending optimization of dry materials and asphalt for carbon anode production, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the steps of the method according to any one of claims 1 to 6.
9. The system for intelligent blending optimization of carbon anode production drier and asphalt according to claim 8, further comprising: The data acquisition unit is used for automatically acquiring the formula parameters and the corresponding physical and chemical performance data of the product from the sensors and the detection equipment on the carbon anode production line; The data interface unit is used for communicating with a production execution system MES or a distributed control system DCS, receiving a production instruction and issuing optimized formula parameters; The system is deployed in an industrial control computer or server.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Description

Method and system for intelligently blending and optimizing dry materials and asphalt in carbon anode production Technical Field The application relates to a method and a system for intelligently blending and optimizing dry materials and asphalt in carbon anode production, and belongs to the technical field of carbon prebaked anode production. Background The carbon prebaked anode (hereinafter referred to as carbon anode) is a key material in the production of electrolytic aluminum, and plays roles of conducting electricity and participating in electrochemical reaction in an electrolytic tank, and the quality of the carbon prebaked anode directly influences current efficiency, energy consumption and anode consumption rate. The production of carbon anodes generally uses petroleum coke as aggregate and coal pitch as binder. The aggregate is crushed and sieved into different particle sizes, such as coarse particle, medium particle, fine particle and powder, and the aggregate is mixed with molten coal asphalt in certain proportion, kneaded, formed and roasted. The granularity composition (i.e. the mass percent of each particle fraction) of the dry material and the addition amount of coal pitch are key technological parameters for determining the final physicochemical properties (such as bulk density, mechanical strength, resistivity and the like) of the carbon anode. In the process of realizing the application, the inventor finds that at least the following problems exist in the prior art, namely the current formula (particle size composition and coal pitch amount) of the carbon anode is mainly subjected to qualitative adjustment according to experience of process personnel, and the law is summarized by observing the performance change of a finished product. The method has low efficiency, is difficult to capture the complex nonlinear coupling relation between the multi-grain grades and the coal tar pitch, cannot systematically find the global optimal formula which enables the multi-performance indexes to reach the optimal simultaneously, and severely restricts the stable improvement of the product quality and the reduction of the production cost. Although the orthogonal experiment method can be used for preliminary screening, the method cannot perform fine optimization in a continuous space, the experiment times increase exponentially with the increase of factors, and the cost is high. Therefore, how to quickly, accurately and low-cost determine the optimal dry material granularity composition and coal pitch ratio based on a plurality of physicochemical performance indexes of the carbon anode becomes a technical problem to be solved in order to realize digital and intelligent transformation in the carbon industry. Disclosure of Invention The application aims to overcome the defects of the prior art, and provides a method, a system and a storage medium for intelligently blending and optimizing dry materials and asphalt in carbon anode production, which can efficiently and accurately find a global optimal formula with minimum experiment times and realize reliable prediction and decision support on the performance of an unknown formula. The technical scheme adopted by the application for solving the technical problems is as follows: in a first aspect, an embodiment of the present application provides a method for optimizing intelligent blending of dry materials and asphalt in carbon anode production, including the following steps: the method comprises the following steps: Step S1, acquiring an initial data set of physicochemical properties of the carbon anode under the proportion of different dry material granularity compositions and coal tar pitch based on orthogonal experimental design, and sequencing the influence importance of all factors, wherein the physicochemical properties at least comprise bulk density, compressive strength, flexural strength and room temperature resistivity; S2, taking the initial data set as a training sample, and constructing a Gaussian process model taking dry material granularity composition and coal asphalt ratio as input and taking physical and chemical properties of a carbon anode as output; Step S3, based on a Bayesian optimization framework, taking an expected lifting EI as an acquisition function, under the guidance of the Gaussian process model, iteratively searching a formula which enables the function of the comprehensive physicochemical property of the carbon anode to be optimal under the constraint that the sum of the percentages of all the granularity components is equal to 100% and the percentages of all the components are within a preset production range until a preset convergence condition is met, wherein the convergence condition comprises that the maximum evaluation times are reached or the relative improvement quantity of the optimal observation value in the continuous k iterations is smaller than a threshold delta; step S4, based on an active learning strategy, per