Graphite purification method and system based on gradient density separation

Abstract

The application provides a graphite purification method and system based on gradient density separation, and relates to the technical field of graphite purification. The method comprises the following steps: measuring the physical properties of the cleaned graphite ore sample; introducing a Bayesian normalization layer to perform liquid gradient concentration distribution noise quantization simulation; traversing a graphite purification scene template library to identify a graphite purification strategy; performing gradient density separation and purification on the graphite ore sample, and combining the physical characteristic information of the graphite ore sample and the liquid gradient concentration distribution monitoring results in the purification process to match and identify the graphite purification scene template library; and obtaining a matched graphite purification strategy template to optimize the gradient density separation and purification process. The application solves the technical problem of unstable purification effect caused by noise interference and parameter uncertainty in the graphite purification process in the prior art, and improves the stability and reliability of the graphite separation effect by quantizing the uncertainty in the graphite purification process through the Bayesian method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of graphite purification, and particularly relates to a graphite purification method and system based on gradient density separation. BACKGROUND

[0002] The purification effect of the existing graphite purification process is highly dependent on the accurate control of key process parameters such as temperature, atmosphere, and reactant concentration. However, in actual industrial production, due to the inherent differences in raw material sources, fluctuations in auxiliary material quality, and significant uncertainty in temperature measurement and control accuracy under high-temperature environment, it is difficult to maintain the key parameters in the purification process to be accurate and consistent, which together constitute continuous noise interference and parameter misalignment, resulting in difficulty in maintaining the best reaction window in the purification process, causing the purity of products in different batches or even the same batch to fluctuate sharply, and seriously affecting the stability of the graphite purification effect.

[0003] In summary, the existing technology has the technical problem of unstable purification effect due to noise interference and parameter uncertainty in the graphite purification process. SUMMARY

[0004] The purpose of the present application is to provide a graphite purification method and system based on gradient density separation to solve the technical problem of unstable purification effect due to noise interference and parameter uncertainty in the graphite purification process in the prior art.

[0005] In view of the above problems, the present application provides a graphite purification method and system based on gradient density separation.

[0006] In a first aspect, the present application provides a graphite purification method based on gradient density separation, which is realized by a graphite purification system based on gradient density separation. The graphite purification method based on gradient density separation comprises: screening and ultrasonic cleaning of graphite ore samples to remove surface impurities, physical property measurement of the cleaned graphite ore samples to obtain physical characteristic information of the graphite ore samples; introducing a Bayesian normalization layer, combining historical graphite purification data to perform liquid gradient concentration distribution noise quantization simulation, and constructing a graphite purification scene template library; traversing the graphite purification scene template library to identify graphite purification strategies, and constructing a graphite purification strategy template library; performing gradient density separation and purification on the graphite ore samples, and combining the physical characteristic information of the graphite ore samples and the liquid gradient concentration distribution monitoring results in the purification process to perform matching identification on the graphite purification scene template library, and obtaining a matching graphite purification scene template; searching the graphite purification strategy template library based on the matching graphite purification scene template to obtain a matching graphite purification strategy template, and optimizing the gradient density separation and purification process.

[0007] Optionally, the historical graphite purification data is data carded according to physical properties, liquid gradient concentration distribution and purification indexes respectively to obtain a historical physical property-liquid gradient concentration distribution-purification index data set; the historical physical property-liquid gradient concentration distribution-purification index data set is aggregated based on physical properties to obtain a plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; noise fluctuation scales are identified according to the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets respectively to obtain a plurality of noise fluctuation scales; the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets are randomly disturbed and expanded based on the plurality of noise fluctuation scales by introducing a Bayesian normalization layer to obtain a plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; a plurality of graphite purification scene template sets are determined by centrally data carding the plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets, and the graphite purification scene template library is constructed by summarizing.

[0008] Optionally, liquid gradient concentration distribution fluctuation analysis is performed on the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to determine a plurality of liquid gradient concentration distribution fluctuation factors; purification index fluctuation analysis is performed on the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to determine a plurality of purification index follow-up factors; noise fluctuation scales are identified based on the plurality of liquid gradient concentration distribution fluctuation factors and the plurality of purification index follow-up factors to obtain a plurality of noise fluctuation scales.

[0009] Optionally, liquid gradient concentration distribution and purification index data in the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets are randomly disturbed according to the plurality of noise fluctuation scales by the Bayesian normalization layer to obtain a plurality of initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets with an approximation greater than a preset approximation threshold in the plurality of initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets are mean processed to obtain the plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets.

[0010] Optionally, Gaussian noise is generated based on the plurality of noise fluctuation scales by the Bayesian normalization layer, and liquid gradient concentration distribution and purification index data in the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets are adjusted according to the generation results respectively to obtain the plurality of initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets.

[0011] Optionally, a pre-constructed graphite purification strategy identifier is identified, wherein the graphite purification strategy identifier comprises an input layer, a convolution layer and an output layer; the graphite purification strategy identifier is used for strategy identification of the graphite purification scene template in the graphite purification scene template library respectively, and the graphite purification strategy template library is obtained.

[0012] Optionally, the graphite purification scene template in the graphite purification scene template library is subjected to cosine similarity calculation respectively with the graphite ore sample physical feature information and the liquid gradient concentration distribution monitoring result in the purification process as indexes, the graphite purification scene template corresponding to the maximum calculation value is extracted, and the matched graphite purification scene template is obtained.

[0013] Optionally, the gradient density separation purification process is continuously monitored to obtain a liquid gradient concentration distribution continuous monitoring result sequence; the liquid gradient concentration distribution continuous monitoring result sequence and the corresponding liquid gradient concentration distribution in the matched graphite purification scene template are subjected to deviation degree analysis to obtain a deviation degree sequence; the deviation degree sequence is analyzed from two dimensions of change trend and centralized data to determine a deviation influence coefficient; and the matched graphite purification strategy template is optimized based on the deviation influence coefficient.

[0014] Optionally, the deviation degree sequence is subjected to centralized data identification by using a mean shift algorithm to obtain a centralized deviation degree; the deviation degree sequence is subjected to change trend feature analysis to obtain a deviation degree change trend feature; and the centralized deviation degree is corrected based on the deviation degree change trend feature to obtain the deviation influence coefficient.

[0015] In a second aspect, the application further provides a graphite purification system based on gradient density separation, used for performing the graphite purification method based on gradient density separation as described in the first aspect, wherein the graphite purification system based on gradient density separation comprises: a physical property measurement module, used for screening graphite ore samples and performing ultrasonic cleaning to remove surface impurities, performing physical property measurement on the cleaned graphite ore samples to obtain physical characteristic information of the graphite ore samples; a noise quantization simulation module, used for introducing a Bayesian normalization layer, performing liquid gradient concentration distribution noise quantization simulation in combination with historical graphite purification data to construct a graphite purification scene template library; a purification strategy identification module, used for identifying graphite purification strategies by traversing the graphite purification scene template library to construct a graphite purification strategy template library; a matching identification module, used for performing gradient density separation purification on the graphite ore samples, and in combination with the physical characteristic information of the graphite ore samples and liquid gradient concentration distribution monitoring results in the purification process, performing matching identification on the graphite purification scene template library to obtain a matching graphite purification scene template; and a strategy retrieval module, used for retrieving the graphite purification strategy template library based on the matching graphite purification scene template to obtain a matching graphite purification strategy template to optimize the gradient density separation purification process.

[0016] The one or more technical solutions provided in the application have at least the following beneficial effects:

[0017] By screening graphite ore samples and performing ultrasonic cleaning to remove surface impurities, performing physical property measurement on the cleaned graphite ore samples to obtain physical characteristic information of the graphite ore samples, introducing a Bayesian normalization layer, performing liquid gradient concentration distribution noise quantization simulation in combination with historical graphite purification data to construct a graphite purification scene template library, traversing the graphite purification scene template library to identify graphite purification strategies to construct a graphite purification strategy template library, performing gradient density separation purification on the graphite ore samples, and in combination with the physical characteristic information of the graphite ore samples and liquid gradient concentration distribution monitoring results in the purification process, performing matching identification on the graphite purification scene template library to obtain a matching graphite purification scene template, and retrieving the graphite purification strategy template library based on the matching graphite purification scene template to obtain a matching graphite purification strategy template to optimize the gradient density separation purification process. That is, by combining the Bayesian method with the gradient density separation technology, the graphite purification process can maintain high stability and reliability in an experimental environment with high uncertainty.

[0018] The above description is only a summary of the technical solutions of the present application. In order to make the technical means of the present application more clear and complete, and to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. It should be understood that the contents described in this part are not intended to identify the key or important features of the embodiments of the present application, nor are they intended to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only exemplary, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0020] Figure 1 The flowchart of the graphite purification method based on gradient density separation of the present application.

[0021] Figure 2 The structural schematic diagram of the graphite purification system based on gradient density separation of the present application.

[0022] Explanation of reference signs: physical property measurement module 11, noise quantization simulation module 12, purification strategy identification module 13, matching identification module 14, strategy retrieval module 15. DETAILED DESCRIPTION

[0023] The present application provides a graphite purification method and system based on gradient density separation, which solves the technical problem of unstable purification effect caused by noise interference and parameter uncertainty in the graphite purification process in the prior art. By combining the Bayesian method with the gradient density separation technology, the graphite purification process can maintain high stability and reliability in an uncertain experimental environment.

[0024] The technical solutions in the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application. In addition, it should be noted that only parts related to the present application are shown in the drawings, not all.

[0025] Embodiment one, please refer to the attached Figure 1The application provides a graphite purification method based on gradient density separation, wherein the graphite purification method based on gradient density separation is performed by a graphite purification system based on gradient density separation, and specifically includes the following steps:

[0026] The graphite ore sample is screened and ultrasonically cleaned to remove surface impurities, and the cleaned graphite ore sample is measured for physical properties to obtain physical characteristic information of the graphite ore sample.

[0027] Specifically, the graphite ore sample is screened to remove impurity particles, and ultrasonic cleaning is used to remove surface-attached impurities. Graphite ore sample screening refers to selecting suitable graphite ore samples for purification from raw ore, and screening according to standards such as particle size, purity, and chemical composition of the graphite ore sample to ensure that the selected graphite ore sample meets the requirements for purification. The carbon content of the graphite ore sample usually needs to meet certain standards to ensure the effect of subsequent purification. For lower quality ore, preliminary physical screening is required. The screened graphite ore sample is placed in an ultrasonic cleaning device and cleaned with deionized water, etc. Through ultrasonic oscillation, dust, grease and other impurities on the surface of the graphite ore sample are removed to ensure that the cleaned graphite ore sample is pure to avoid the influence of impurities on subsequent physical property testing and purification process. For example, in the ultrasonic cleaning experiment, an ultrasonic oscillation frequency of 40-50 kHz is used, and the cleaning time is 15-20 minutes. During this process, the removal rate of impurities on the surface of the graphite ore sample can usually reach more than 90%.

[0028] The cleaned graphite ore sample is further tested for physical properties to obtain physical characteristic information (such as density, particle size, etc.) of the graphite ore sample. Specific surface area and particle size distribution are usually tested using a BET specific surface area analyzer and a laser particle size analyzer, and density can be measured using a gas adsorption method or a water displacement method. For example, assuming that the specific surface area of the cleaned graphite ore sample is 25 m² / g and the particle size distribution is 10-50 μm.

[0029] A Bayesian normalization layer is introduced to combine historical graphite purification data to simulate noise quantization of liquid gradient concentration distribution and construct a graphite purification scenario template library.

[0030] Further, the application further comprises the following steps: data carding the historical graphite purification data according to physical properties, liquid gradient concentration distribution and purification indexes respectively to obtain a historical physical property-liquid gradient concentration distribution-purification index data set; homogenizing the historical physical property-liquid gradient concentration distribution-purification index data set based on physical properties to obtain a plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; identifying noise fluctuation scales according to the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets respectively to obtain a plurality of noise fluctuation scales; introducing a Bayesian normalization layer to randomly perturb and expand the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets according to the plurality of noise fluctuation scales to obtain a plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; and determining a plurality of graphite purification scene template sets by centrally carding the plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets, and constructing the graphite purification scene template library by summarizing.

[0031] Further, the application further comprises the following steps: liquid gradient concentration distribution fluctuation analysis on the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to determine a plurality of liquid gradient concentration distribution fluctuation factors; purification index fluctuation analysis on the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to determine a plurality of purification index following factors; and noise fluctuation scale identification based on the plurality of liquid gradient concentration distribution fluctuation factors and the plurality of purification index following factors to obtain a plurality of noise fluctuation scales.

[0032] Further, the application further comprises the following steps: randomly perturbing liquid gradient concentration distribution and purification index data in the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets according to the plurality of noise fluctuation scales respectively by the Bayesian normalization layer to obtain a plurality of initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; and mean processing initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data with an approximation greater than a preset approximation threshold in the plurality of initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to obtain the plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets.

[0033] Further, the application further comprises the following steps: generating Gaussian noise based on the plurality of noise fluctuation scales through the Bayesian normalization layer, and adjusting the liquid gradient concentration distribution and the purification index data in the plurality of initial extended aggregated historical physical property-liquid gradient concentration distribution-purification index data sets according to the generated results, to obtain the plurality of initial extended aggregated historical physical property-liquid gradient concentration distribution-purification index data sets.

[0034] Specifically, historical graphite purification experimental data is collected, including the physical properties of graphite ore samples (such as density, particle size, and morphology) and experimental parameters (such as liquid gradient concentration, centrifugal force, and temperature) under different experimental conditions. The historical graphite purification data includes experimental results (such as graphite purity, purification efficiency, and impurity removal rate) and specific values of various parameters under different purification methods. The collected historical graphite purification data is cleaned to remove missing values or inconsistent data. The cleaned historical graphite purification data is standardized to ensure that the data under different experimental conditions are comparable and provide a unified input format for subsequent Bayesian modeling.

[0035] Physical properties are basic physical parameters of graphite ore samples, such as particle size, specific surface area, density, and electrical conductivity; during graphite purification, the concentration of the liquid may have a gradient distribution in different regions, affecting the separation efficiency of graphite, and the liquid gradient concentration distribution is the distribution of the liquid concentration in space, which directly affects the reaction and purification effect of the graphite ore sample during purification; purification indicators are indicators used to measure the purity or purification efficiency of graphite during purification, such as graphite purity, impurity removal rate, and purification efficiency.

[0036] According to the physical properties, liquid gradient concentration distribution, and purification indicators, the historical graphite purification data is sorted and organized into a unified set, forming a historical physical property-liquid gradient concentration distribution-purification indicator data set. The historical physical property-liquid gradient concentration distribution-purification indicator data set is a composite data set containing physical properties, liquid gradient concentration distribution, and purification indicators. Each data set consists of three parts: physical properties, liquid gradient concentration distribution, and purification indicators. For example, a complete historical data set contains physical properties: particle size range (10-50 μm), specific surface area (25 m² / g), and density (2.3 g / cm³); liquid gradient concentration distribution increases from 10% solvent concentration to 50%; purification indicators: graphite purity increases from 55% to 85%, and impurity removal rate is 45%.

[0037] According to the physical characteristics of the graphite ore sample, the historical physical characteristics-liquid gradient concentration distribution-purification index data set is aggregated in the same category, and the data sets with similar characteristics are aggregated according to the physical characteristics to form multiple aggregated historical physical characteristics-liquid gradient concentration distribution-purification index data sets. The purpose of the same aggregation is to find a group of data with similar physical characteristics.

[0038] In the graphite purification process, due to the fluctuation of process parameters and external interference, it may cause uncertainty or noise fluctuation in the purification process. The multiple aggregated historical physical characteristics-liquid gradient concentration distribution-purification index data sets are respectively subjected to noise fluctuation scale identification, and the amplitude and range of noise fluctuation are quantified.

[0039] Specifically, the multiple aggregated historical physical characteristics-liquid gradient concentration distribution-purification index data sets are traversed to perform liquid gradient concentration distribution fluctuation analysis, analyze the liquid gradient concentration distribution fluctuation in the purification process of different graphite ore samples, and determine multiple liquid gradient concentration distribution fluctuation factors. In the graphite purification process, the concentration of the liquid changes at different positions and times usually shows a gradient distribution, and the liquid gradient concentration distribution fluctuation analysis refers to the statistics and analysis of the changes of these concentration gradients, identifies the fluctuation law of the liquid concentration distribution, and understands the influence of the concentration fluctuation on the purification effect. The liquid gradient concentration distribution fluctuation factor refers to the specific characteristics or reasons of the liquid concentration change. For example, temperature change, liquid flow speed or solubility, etc. The identification of the fluctuation factor helps to understand which factors affect the liquid concentration fluctuation.

[0040] Similarly, the multiple aggregated historical physical characteristics-liquid gradient concentration distribution-purification index data sets are traversed to perform purification index fluctuation analysis, and the fluctuation of the purification index is usually related to multiple factors in the purification process, especially the liquid concentration, the physical characteristics of the graphite ore sample, the temperature, etc. Multiple purification index follow-up factors are determined. The analysis of the fluctuation of each index (such as graphite purity, impurity removal rate, etc.) in the purification process helps to identify the key factors affecting the purification effect. The purification index fluctuates with the fluctuation of the purification parameter. The purification index fluctuation analysis refers to the statistical analysis of the fluctuation of the related index (such as graphite purity, impurity removal rate, etc.) in the purification process, and identifies which factors or conditions cause the fluctuation of the purification effect. The purification index follow-up factor is a dynamic variable related to the fluctuation factor in the purification process, such as temperature, liquid concentration, particle size, etc. The purification index follow-up factor can reflect the dynamic conditions affecting the index fluctuation in the purification process.

[0041] Based on the liquid gradient concentration distribution fluctuation factor and the purification index follow-up factor, noise fluctuation scale identification is performed to quantify the amplitude and range of the fluctuations, and then determine the degree of influence of these fluctuations on the purification process. Noise fluctuation scale identification is performed by analyzing the fluctuations in liquid concentration and purification index to determine their amplitude and influence range. By identifying the noise fluctuation scale, errors caused by unstable factors during the purification process can be predicted and reduced, thereby improving the stability of the purification process. By analyzing the error range in the historical data, appropriate noise intensity and fluctuation scale are determined. Different noise scales can simulate different types of disturbances in the experiment, such as changes in liquid concentration, temperature fluctuations, fluctuations in centrifugal force, etc.

[0042] The liquid gradient concentration distribution and purification index data in the multiple aggregated historical physical property-liquid gradient concentration distribution-purification index data sets are processed by the Bayesian normalization layer. The liquid gradient concentration distribution and purification index data are randomly disturbed by the Bayesian normalization layer, that is, a certain randomness is introduced on the data to simulate the uncertainty in the real world. According to the fluctuation characteristics of the historical data, especially the noise fluctuation scale (such as the fluctuation amplitude of liquid concentration or the change amplitude of purification index), noise conforming to Gaussian distribution is generated. The process of noise generation is based on the previous analysis of data fluctuation. The mean and standard deviation of Gaussian noise match the fluctuation in the historical data. Once the Gaussian noise is generated, the liquid gradient concentration distribution and purification index data in the multiple aggregated historical physical property-liquid gradient concentration distribution-purification index data sets can be adjusted according to the generation result of the Gaussian noise. The adjusted data will contain a certain randomness to simulate the uncertainty that may occur in reality. The Bayesian normalization layer considers the uncertainty of the data by calculating the mean and variance of the input data and performs normalization processing to ensure the consistency of the adjusted data with the original data and avoid excessive disturbance of the data. The result of normalization will make the liquid gradient concentration distribution data and the purification index data change within a suitable range while maintaining the relative relationship and operability of the data. After the Bayesian normalization layer completes the normalization processing, the liquid gradient concentration distribution and the purification index data will be adjusted into the expanded initial data set, simulating more purification scenarios and providing rich test data for subsequent optimization. All the expanded data sets generated by random disturbance are integrated with the original historical data set to form a complete training set containing diversified data, that is, multiple initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets. The expanded data set can simulate the purification effect under different fluctuation and uncertainty conditions. For example, in some scenarios, the fluctuation amplitude of liquid concentration is large, and the purity may fluctuate greatly; in other scenarios, the fluctuation is small, and the purification effect is stable. Through the expanded data set, the model can learn the purification strategy under different conditions, thereby improving its adaptability. For example, assume that there are 3 different sources of graphite ore sample data sets in the example, including the physical properties of the graphite ore samples (such as particle size, specific surface area, etc.), the liquid concentration (fluctuating between 20%-50%), and the corresponding purification index (graphite purity and impurity removal rate). The historical data set: Experiment 1 liquid concentration 30%, purity 80%, impurity removal rate 50%; Experiment 2 liquid concentration 40%, purity 82%, impurity removal rate 52%; Experiment 3 liquid concentration 50%, purity 85%, impurity removal rate 55%. Through fluctuation analysis, it is found that the fluctuation amplitude of liquid concentration is ±5%, and the fluctuation amplitude of purification index is ±2% purity. By generating noise conforming to Gaussian distribution, the random fluctuation of liquid concentration and purification index is simulated.The Bayesian normalization adjusts these Gaussian noises, ensuring that the data varies within a reasonable range. For example, with a liquid concentration of 30%, the generated purity data may be adjusted from 80% to between 78%-82%. The expanded data set: Experiment 4 liquid concentration 30%, purity 79%, impurity removal rate 51%; Experiment 5 liquid concentration 40%, purity 83%, impurity removal rate 53%; Experiment 6 liquid concentration 50%, purity 84%, impurity removal rate 56%.

[0043] Bayesian normalization is a normalization method based on Bayesian inference, which processes data by considering the uncertainty of the data. In data expansion and noise analysis, the Bayesian normalization layer can make reasonable adjustments and processing according to the noise characteristics in historical data, thereby improving the robustness and prediction ability of the model. Through the Bayesian normalization layer, multiple aggregated historical physical property-liquid gradient concentration distribution-purification index data sets are randomly disturbed, that is, a certain randomness is introduced on the data set to expand the data set, to simulate the uncertainty in the real world. Random disturbance helps to simulate the noise and uncertainty in the data, so that the model can more accurately cope with complex situations in the real world.

[0044] Approximation analysis is performed on multiple initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to measure the similarity between different data sets. When the approximation degree of some data sets is greater than the preset threshold, it means that they are very similar in some aspects. Approximation degree refers to the similarity between data points, which can be measured by calculating the distance or similarity between data points. A preset approximation threshold is set, such as 0.8. The mean value of the initial expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data with an approximation degree greater than the preset approximation threshold is processed, and the average value of these similar data points is calculated to reduce the influence of noise and outliers, thereby obtaining a more stable and reliable data set. Mean value processing refers to averaging the values of multiple data sets to reduce the volatility between data and extract more stable trends, which is used to adjust the expanded data with an approximation degree greater than the threshold, to ensure the stability of these data sets and reduce unnecessary fluctuations. For example, assuming that the approximation degree of graphite ore sample 1 and graphite ore sample 2 is greater than the preset threshold (such as 0.85), the mean value of these two data sets is processed: Experiment 1 and Experiment 2 mean value: liquid concentration 32.5%, purity 79%, impurity removal rate 51%.

[0045] After mean processing, the final expanded aggregated historical physical properties-liquid gradient concentration distribution-purification index data set is obtained, which contains historical physical properties, liquid gradient concentration distribution and purification index data, reflecting the influence of changes in different process parameters (such as liquid concentration, temperature, etc.) on the purification effect, and also helping to improve the accuracy and robustness of the purification process. Through random disturbance and mean processing, the data set is expanded to simulate the purification effect under different fluctuation conditions, and through mean processing, the fluctuation in the data is reduced, and the stability and reliability of the data are enhanced.

[0046] By centrally sorting multiple expanded aggregated historical physical properties-liquid gradient concentration distribution-purification index data sets, effective data is selected and redundant or irrelevant parts are removed. For example, if some data repeatedly appears in multiple experiments, it can be summarized as a representative value to avoid repeated calculation. Ensure that all data is organized according to a unified standard format, and all data items (such as liquid concentration, graphite purity, etc.) need to be converted to a unified unit of measurement to avoid errors caused by inconsistent units. From the centrally sorted data, extract the key features that affect the purification effect of graphite, including the physical properties of graphite ore samples (such as particle size, specific surface area, density, etc.), liquid gradient concentration distribution (such as solvent concentration variation), and indicators during the purification process (such as purity, impurity removal rate, etc.). Each type of data is aggregated into an independent graphite purification scenario template, including different liquid concentrations, ore sample characteristics, purification indicators, etc., reflecting the graphite purification effect under different process conditions.

[0047] After the above process, multiple graphite purification scenario templates are obtained, covering different purification environments, process parameters and purification effects. All the purification scenario templates are collected to form a complete graphite purification scenario template library. The graphite purification scenario template library is a database formed by the collection of multiple purification scenario templates, which is a complete template set under different purification conditions, and can help choose and optimize purification strategies in actual production process. Through the sorting and aggregation of historical data, efficient use of data is ensured, and the influence of uncertainty on the purification effect is reduced through Bayesian normalization. By generating the expanded data set and the graphite purification scenario template library, the purification process under different conditions can be simulated to help choose the most suitable purification strategy.

[0048] Iterate through the graphite purification scenario template library to identify graphite purification strategies and build a graphite purification strategy template library.

[0049] Further, the present application further comprises the following steps: pre-constructing a graphite purification strategy identifier, wherein the graphite purification strategy identifier comprises an input layer, a convolution layer and an output layer; using the graphite purification strategy identifier to respectively identify the graphite purification strategy of the graphite purification scene template in the graphite purification scene template library, and obtaining the graphite purification strategy template library.

[0050] Specifically, the graphite purification strategy identifier is a machine learning-based method designed to identify and select appropriate purification strategies based on input data such as physical properties of graphite ore samples, liquid gradient concentration, purification indicators, etc. to predict or select the best purification strategy. The graphite purification strategy identifier learns the key features in the purification process through training data, and gives the optimal purification strategy through the output layer. To build a graphite purification strategy identifier, the graphite purification strategy identifier is based on a convolutional neural network (CNN) architecture. The graphite purification strategy identifier includes an input layer, a convolution layer and an output layer. The input layer receives features from historical graphite purification data, including physical properties of graphite ore samples (such as particle size, specific surface area, density, etc.), liquid gradient concentration distribution and purification indicators (such as graphite purity, impurity removal rate, etc.). The convolution layer is used to extract features from the input data. The convolution layer scans the input data through filters to identify patterns and trends that affect purification results. For example, the convolution layer identifies the impact of liquid concentration fluctuations and changes in sample physical properties on purification results. The output layer is used to output the final purification strategy based on the network's calculation results, including the setting of liquid concentration, temperature control, reaction time and other parameters.

[0051] A large number of graphite purification sample data is collected and organized, including physical properties of graphite ore samples, liquid gradient concentration distribution and purification indicators, etc. The graphite purification sample data is divided into 80% training set and 20% validation set. The training set is used to train the graphite purification strategy identifier. During the training process, the graphite purification strategy identifier tries to match the input data and the corresponding purification strategy. Through optimization algorithm, the parameters of the graphite purification strategy identifier are constantly adjusted to minimize the value of the loss function. During the training process, the validation set is used regularly to evaluate the performance of the graphite purification strategy identifier to ensure that the graphite purification strategy identifier does not overfit and can generalize to new, unseen data. Through training, the graphite purification strategy identifier can learn the complex mapping relationship from the graphite purification scene to the purification strategy.

[0052] The trained graphite purification strategy recognizer is used to identify the strategy of each template in the graphite purification scene template library. By inputting the data such as the characteristics of the ore sample, the liquid gradient concentration distribution, and the purification index in each purification scene template, the graphite purification strategy recognizer will assign a suitable purification strategy to each scene template according to the existing learning mode. For example, assuming that a template contains a graphite ore sample with a liquid concentration range from 30% to 50%, the graphite purification strategy recognizer will analyze the data of this template (such as particle size, fluctuation range of liquid concentration, etc.) and select the most suitable purification strategy, including increasing the liquid concentration, adjusting the temperature, etc.

[0053] Through the identification of each graphite purification scene template, a corresponding purification strategy template is generated, forming a new graphite purification strategy template library that summarizes the optimized purification strategies selected based on different ore sample characteristics and process conditions, including strategies for multiple purification processes, suitable for different graphite ore samples and production conditions. Through the pre-constructed graphite purification strategy recognizer, effective purification strategies are automatically identified and extracted from the graphite purification scene template library, improving the efficiency of formulating purification strategies and reducing the cost of manual analysis and experiments.

[0054] The graphite ore sample is subjected to gradient density separation and purification, and the graphite purification scene template library is matched and identified based on the physical characteristics information of the graphite ore sample and the monitoring results of the liquid gradient concentration distribution during the purification process, to obtain a matched graphite purification scene template.

[0055] Further, the present application further includes the following steps: using the physical characteristics information of the graphite ore sample and the monitoring results of the liquid gradient concentration distribution during the purification process as an index, respectively calculating the cosine similarity of the graphite purification scene templates in the graphite purification scene template library, extracting the graphite purification scene template corresponding to the maximum calculation value, and obtaining the matched graphite purification scene template.

[0056] Specifically, the graphite ore sample is subjected to gradient density separation and purification. Gradient density separation is a physical method commonly used in mineral purification, which utilizes the density difference of different substances in liquid to separate minerals through a liquid medium with gradually changing density. In the graphite purification process, the graphite ore sample is added to a liquid with gradient density, and the graphite is separated by density difference to remove impurities. The graphite ore sample is put into a liquid medium with gradient density, and the liquid concentration has a gradient between different levels, with gradually increasing density. The density difference causes different mineral particles to sink or float in the liquid according to their density, and finally separates the graphite and impurities. The graphite ore sample is layered in the liquid according to the density difference, and the graphite usually stays in a certain density region, while the impurities are excluded.

[0057] The physical characteristic information of the graphite ore sample is the basic physical properties of the graphite ore sample, such as particle size, specific surface area, density, hardness, etc. In the process of graphite purification, the distribution and change of liquid concentration will affect the purification effect. The monitoring result of liquid gradient concentration distribution is the change of liquid concentration under different purification conditions, which is used to analyze the concentration fluctuation in the purification process and its influence on the purification effect of the ore sample. The physical characteristic information of the graphite ore sample and the monitoring result of liquid gradient concentration distribution in the purification process are used as indexes, that is, these data are used as input features for comparison with each template in the graphite purification scene template library.

[0058] By calculating the cosine similarity, the physical characteristic information of the graphite ore sample and the monitoring result of liquid gradient concentration distribution in the purification process are respectively calculated with the graphite purification scene template in the graphite purification scene template library to obtain a similarity value, which represents the similarity between the current graphite ore sample and each graphite purification scene template in the graphite purification scene template library. The greater the similarity, the higher the similarity between the current graphite ore sample and the template, and the more suitable it is to use the purification strategy of the template. The graphite purification scene template corresponding to the maximum cosine similarity value calculated is extracted as the best matching template of the current ore sample. For example, in an example, the particle size of graphite ore sample A is 20-50 μm, the liquid concentration is 30%-50%, the graphite purity is 80%, and the impurity removal rate is 50%. The templates in the graphite purification scene template library are as follows: template 1: particle size 20-50 μm, liquid concentration 30%-50%, graphite purity 80%, and impurity removal rate 50%; template 2: particle size 10-40 μm, liquid concentration 40%-60%, graphite purity 85%, and impurity removal rate 55%; template 3: particle size 50-100 μm, liquid concentration 20%-40%, graphite purity 75%, and impurity removal rate 45%. The similarity between sample A and template 1 is 1.0 (complete match); the similarity with template 2 is 0.85 (similar but not completely matched); and the similarity with template 3 is 0.6 (lower similarity). The maximum cosine similarity is template 1, so template 1 is extracted.

[0059] By matching with all the graphite purification scene templates in the graphite purification scene template library, the most suitable purification strategy for the current ore sample can be ensured, the purification effect is optimized, and experimental errors and fluctuations are reduced. Cosine similarity calculation can help identify the most similar purification strategy template to the current ore sample, so that the production conditions can be adjusted according to the process parameters (such as liquid concentration, reaction time, etc.) in the template, the process parameters are optimized, and the graphite purity and purification efficiency are improved.

[0060] Based on the matched graphite purification scene template, the graphite purification strategy template library is searched to obtain a matched graphite purification strategy template, which optimizes the gradient density separation and purification process.

[0061] Further, the application further comprises the following steps: continuously monitoring the gradient density separation purification process to obtain a sequence of continuous monitoring results of the liquid gradient concentration distribution; analyzing the sequence of continuous monitoring results of the liquid gradient concentration distribution for deviation degree with the corresponding liquid gradient concentration distribution in the matching graphite purification scene template to obtain a sequence of deviation degrees; analyzing the sequence of deviation degrees from two dimensions of change trend and concentrated data to determine a deviation influence coefficient; and optimizing the matching graphite purification strategy template based on the deviation influence coefficient.

[0062] Further, the application further comprises the following steps: identifying concentrated data of the sequence of deviation degrees using a mean shift algorithm to obtain a concentrated deviation degree; analyzing the change trend characteristics of the sequence of deviation degrees to obtain deviation degree change trend characteristics; and correcting the concentrated deviation degree based on the deviation degree change trend characteristics to obtain the deviation influence coefficient.

[0063] Specifically, according to the matching graphite purification scene template, a search is performed in the graphite purification strategy template library to obtain a matching graphite purification strategy template that is most suitable for the matching graphite purification scene template, which contains corresponding purification process parameters such as liquid concentration, temperature, and reaction time. That is, the purification strategy template that is most suitable for the current scene is selected from the graphite purification strategy template library.

[0064] In the gradient density separation purification process, the gradient concentration distribution in the liquid is monitored in real time by a sensor, reflecting the concentration changes in different regions of the liquid. The gradient concentration distribution data of the liquid is continuously recorded to form a sequence of continuous monitoring results, reflecting the changes in the liquid concentration during the purification process. In the gradient density separation purification process, the liquid concentration changes with time and depth. By continuously monitoring the gradient concentration distribution of the liquid, the data sequence of the concentration changes is recorded to obtain the sequence of continuous monitoring results of the liquid gradient concentration distribution.

[0065] By analyzing the sequence of continuous monitoring results of the liquid gradient concentration distribution for deviation degree with the corresponding liquid gradient concentration distribution in the matching graphite purification scene template, the difference between the actual measurement results and the ideal results is analyzed, and the sequence of deviation degrees reflects the difference between the actual liquid concentration and the corresponding liquid gradient concentration distribution in the graphite purification scene template. That is, for each liquid gradient concentration distribution monitoring result at a time, it is compared with the corresponding liquid gradient concentration distribution, and the deviation degree is calculated. The deviation degree is usually the difference between two data, represented as a numerical value. For example, if the corresponding liquid gradient concentration in the graphite purification scene template is 40%, and the actual monitoring value is 38%, the deviation degree is 2%. If the monitoring value is 42%, the deviation degree is 2%.

[0066] By comparing the monitoring data and the template data at all times, the deviation degree at each time is calculated, and these deviation degrees are arranged in time sequence to generate a deviation degree sequence. For example, the deviation degree at time point 1 is 0.5%, the deviation degree at time point 2 is 1.2%, the deviation degree at time point 3 is 0.8%, and the deviation degree at time point 4 is 1.0%.

[0067] By analyzing the deviation degree sequence, the stability of the liquid concentration fluctuation in the purification process can be evaluated. Change trend analysis can reveal whether the deviation degree is gradually increasing, gradually decreasing, or remaining stable. If the deviation degree is small, it means that the liquid concentration changes within the ideal range, and the purification process is relatively stable; if the deviation degree is large, it means that the liquid concentration fluctuates greatly during the purification process, which may lead to unstable purification results and needs to be optimized.

[0068] The mean shift algorithm is used to identify the concentrated data of the deviation degree sequence, and the concentrated area of the deviation degree is identified, which represents the stable stage of the liquid concentration deviation during the purification process. The deviation degree sequence is input, and for each data point in the deviation degree sequence, the mean shift algorithm will move to the area with higher data density according to the density of the surrounding data points until it reaches the maximum density point, identifying the concentrated area of the deviation degree in the sequence. The mean shift algorithm analyzes the data density of the deviation degree sequence and identifies the stable stage of the deviation degree (i.e., the area with smaller fluctuations). Time series analysis is performed on the deviation degree sequence to analyze the trend of the deviation degree sequence over time, identify the change pattern of the deviation degree, and identify whether the deviation degree is gradually increasing, gradually decreasing, or remaining stable, and extract the specific change trend features. For example, the deviation degree in some periods may increase due to external factors (such as temperature changes), or the deviation degree may remain stable due to good control of the liquid concentration.

[0069] The bias influence coefficient is calculated based on the change trend of the bias degree. The bias degree is corrected according to the change trend analysis result. For example, if the bias degree gradually increases at a certain stage, the centralized bias degree can be adjusted to increase its weight to reflect the gradual influence of the bias on the purification effect. According to the corrected centralized bias degree, the bias influence coefficient is calculated. The bias influence coefficient is usually determined according to the size of the corrected centralized bias degree and the intensity of the change trend. The larger the bias influence coefficient, the stronger the influence of the bias degree on the purification effect; on the contrary, the smaller the bias influence coefficient, the weaker the influence of the bias on the purification effect. For example, assume that there is a graphite purification experiment, the liquid concentration range is 30%-50%, and the change data of the liquid concentration in the purification process is obtained through the monitoring equipment, and the bias degree sequence [0.5%, 1.2%, 0.8%, 1.0%, 0.7%] is generated, which represents the bias degree at different time points. Through the mean shift algorithm, the bias degree sequence is analyzed and the centralized region is identified, and the region where the bias degree fluctuates between 0.5%-1.0% is considered as a centralized region, and the centralized bias degree of the region is calculated as 0.8%. Through time series analysis, it is found that the bias degree presents a gradually increasing trend in the first half of the experiment (0.5% to 1.2%), and remains relatively stable in the second half (1.0% to 0.7%). The change trend of the bias degree is first increasing and then stable. According to the change trend analysis, the centralized bias degree is corrected to 0.9% to reflect the influence of the stable bias degree in the later stage of the experiment. According to the corrected centralized bias degree and the change trend, the bias influence coefficient is calculated as 0.6, indicating that the influence of the bias on the purification effect is moderate.

[0070] According to the deviation influence coefficient, the matching graphite purification strategy template is optimized, the liquid concentration range, temperature control and reaction time are adjusted, the optimized matching graphite purification strategy template is obtained, and the process parameters such as liquid concentration, reaction time and temperature are set. The fluctuation of liquid concentration is monitored in real time, and the liquid concentration distribution in the purification process is adjusted according to the deviation degree analysis result, so that the stability of the purification process is ensured. Through the analysis of the change trend of the deviation degree, the key factors affecting the purification effect can be identified, and the matching graphite purification strategy template is optimized, that is, the process parameters in the matching graphite purification strategy template are adjusted. For example, assuming that the liquid concentration of graphite ore sample C is 40%-50%, the purification result is that the purity of graphite is 90% and the impurity removal rate is 60%; the gradient density separation purification process of the graphite ore sample C is continuously monitored, and the monitoring data obtained is [40%, 45%, 48%, 50%, 49%]; the matched purification scene template is the liquid concentration range 40%-50%. The ideal concentration range matched with the template is 40%-50%, the concentration fluctuation in the monitoring data is [40%, 45%, 48%, 50%, 49%], and therefore the deviation degree sequence is [0%, 0%, 2%, 0%, 1%]. The deviation degree is small, indicating that the liquid concentration control is accurate. The deviation influence coefficient is small, indicating that the purification process is very stable in terms of liquid concentration control. Through matching the most suitable graphite purification scene template and deviation degree analysis, the most suitable purification strategy for the current ore sample can be selected, and the purification process is stable and efficient.

[0071] In summary, the graphite purification method based on gradient density separation provided in the present application has the following beneficial effects:

[0072] The graphite ore sample is screened and ultrasonic cleaning is performed to remove surface impurities, the physical property of the cleaned graphite ore sample is measured to obtain the physical characteristic information of the graphite ore sample; a Bayesian normalization layer is introduced, the liquid gradient concentration distribution noise is quantitatively simulated by combining historical graphite purification data, and a graphite purification scene template library is constructed; graphite purification strategy recognition is performed by traversing the graphite purification scene template library, and a graphite purification strategy template library is constructed; the graphite ore sample is subjected to gradient density separation and purification, and the graphite purification scene template library is matched and recognized by combining the physical characteristic information of the graphite ore sample and the liquid gradient concentration distribution monitoring result in the purification process, to obtain a matched graphite purification scene template; the graphite purification strategy template library is searched based on the matched graphite purification scene template, to obtain a matching graphite purification strategy template, and the gradient density separation and purification process is optimized. That is, by combining the Bayesian method with the gradient density separation technology, the graphite purification process can maintain high stability and reliability in an experimental environment with high uncertainty.

[0073] Embodiment two, based on the same inventive concept as the aforementioned embodiment one, a graphite purification system based on gradient density separation is also provided, please refer to the attached Figure 2 , the graphite purification system based on gradient density separation comprises:

[0074] The physical property measurement module 11 is used for screening and ultrasonic cleaning of the graphite ore sample to remove surface impurities, and measuring the physical properties of the cleaned graphite ore sample to obtain physical characteristic information of the graphite ore sample; the noise quantization simulation module 12 is used for introducing a Bayesian normalization layer, combining historical graphite purification data to perform noise quantization simulation of liquid gradient concentration distribution, and constructing a graphite purification scene template library; the purification strategy identification module 13 is used for traversing the graphite purification scene template library to identify graphite purification strategies, and constructing a graphite purification strategy template library; the matching identification module 14 is used for gradient density separation and purification of the graphite ore sample, and combining the physical characteristic information of the graphite ore sample and the liquid gradient concentration distribution monitoring results in the purification process to perform matching identification on the graphite purification scene template library to obtain a matching graphite purification scene template; the strategy retrieval module 15 is used for retrieving the graphite purification strategy template library based on the matching graphite purification scene template to obtain a matching graphite purification strategy template to optimize the gradient density separation and purification process.

[0075] Further, the noise quantization simulation module 12 in the graphite purification system based on gradient density separation is also used for: respectively according to physical properties, liquid gradient concentration distribution and purification indexes, data carding the historical graphite purification data to obtain a historical physical property-liquid gradient concentration distribution-purification index data set; based on physical properties, the historical physical property-liquid gradient concentration distribution-purification index data set is aggregated in the same category to obtain a plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; according to a plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets, noise fluctuation scales are identified respectively to obtain a plurality of noise fluctuation scales; a Bayesian normalization layer is introduced to randomly perturb and expand the plurality of noise fluctuation scales based on the plurality of noise fluctuation scales to obtain a plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets; by centrally carding the plurality of expanded aggregated historical physical property-liquid gradient concentration distribution-purification index data sets, a plurality of graphite purification scene template sets are determined, and the graphite purification scene template library is constructed.

[0076] Further, the noise quantization simulation module 12 in the graphite purification system based on gradient density separation is further used for: traversing the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to perform liquid gradient concentration distribution fluctuation analysis and determine a plurality of liquid gradient concentration distribution fluctuation factors; traversing the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to perform purification index fluctuation analysis and determine a plurality of purification index follow-up factors; and performing noise fluctuation scale identification based on the plurality of liquid gradient concentration distribution fluctuation factors and the plurality of purification index follow-up factors to obtain a plurality of noise fluctuation scales.

[0077] Further, the noise quantization simulation module 12 in the graphite purification system based on gradient density separation is further used for: traversing the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to perform liquid gradient concentration distribution fluctuation analysis and determine a plurality of liquid gradient concentration distribution fluctuation factors; traversing the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to perform purification index fluctuation analysis and determine a plurality of purification index follow-up factors; and performing noise fluctuation scale identification based on the plurality of liquid gradient concentration distribution fluctuation factors and the plurality of purification index follow-up factors to obtain a plurality of noise fluctuation scales.

[0078] Further, the noise quantization simulation module 12 in the graphite purification system based on gradient density separation is further used for: traversing the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to perform liquid gradient concentration distribution fluctuation analysis and determine a plurality of liquid gradient concentration distribution fluctuation factors; traversing the plurality of aggregated historical physical property-liquid gradient concentration distribution-purification index data sets to perform purification index fluctuation analysis and determine a plurality of purification index follow-up factors; and performing noise fluctuation scale identification based on the plurality of liquid gradient concentration distribution fluctuation factors and the plurality of purification index follow-up factors to obtain a plurality of noise fluctuation scales.

[0079] Further, the purification strategy identification module 13 in the graphite purification system based on gradient density separation is further used for: pre-constructing a graphite purification strategy identifier, wherein the graphite purification strategy identifier includes an input layer, a convolution layer, and an output layer; and using the graphite purification strategy identifier to perform strategy identification on the graphite purification scene templates in the graphite purification scene template library respectively to obtain the graphite purification strategy template library.

[0080] Further, the matching recognition module 14 in the graphite purification system based on gradient density separation is further used for: taking the graphite ore sample physical feature information and the liquid gradient concentration distribution monitoring result in the purification process as indexes, respectively performing cosine similarity calculation on the graphite purification scene template in the graphite purification scene template library, extracting the graphite purification scene template corresponding to the maximum calculation value, and obtaining a matched graphite purification scene template.

[0081] Further, the strategy retrieval module 15 in the graphite purification system based on gradient density separation is further used for: continuously monitoring the gradient density separation purification process to obtain a liquid gradient concentration distribution continuous monitoring result sequence; performing deviation degree analysis on the liquid gradient concentration distribution continuous monitoring result sequence and the corresponding liquid gradient concentration distribution in the matched graphite purification scene template to obtain a deviation degree sequence; analyzing the deviation degree sequence from two dimensions of change trend and centralized data to determine a deviation influence coefficient; and optimizing the matched graphite purification strategy template based on the deviation influence coefficient.

[0082] Further, the strategy retrieval module 15 in the graphite purification system based on gradient density separation is further used for: performing centralized data identification on the deviation degree sequence by using a mean shift algorithm to obtain a centralized deviation degree; performing change trend feature analysis on the deviation degree sequence to obtain a deviation degree change trend feature; and correcting the centralized deviation degree based on the deviation degree change trend feature to obtain the deviation influence coefficient.

[0083] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The foregoing Figure 1 The graphite purification method based on gradient density separation in embodiment one and the specific examples are also applicable to the graphite purification system based on gradient density separation in the present embodiment. Through the foregoing detailed description of the graphite purification method based on gradient density separation, those skilled in the art can clearly know the graphite purification system based on gradient density separation in the present embodiment. Therefore, for the sake of brevity of the specification, no further detailed description is given here.

[0084] The above description of the disclosed embodiments enables a person 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 can 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.

[0085] Obviously, for those skilled in the art, some improvements and modifications can be made to the present application without departing from the principles of the present application, and these improvements and modifications also fall within the protection scope of the present application.

Claims

1. A graphite purification method based on gradient density separation, characterized in that, include: Graphite ore samples were screened and ultrasonically cleaned to remove surface impurities. The physical properties of the cleaned graphite ore samples were then measured to obtain physical characteristic information of the graphite ore samples. A Bayesian normalization layer is introduced, and historical graphite purification data is combined to simulate the noise quantization of liquid gradient concentration distribution, thereby constructing a graphite purification scenario template library. The graphite purification scenario template library is traversed to identify graphite purification strategies, and a graphite purification strategy template library is constructed. The graphite ore sample is subjected to gradient density separation and purification. Based on the physical characteristics of the graphite ore sample and the monitoring results of the liquid gradient concentration distribution during the purification process, the graphite purification scenario template library is matched and identified to obtain a matching graphite purification scenario template. Based on the matching graphite purification scenario template, the graphite purification strategy template library is searched to obtain the matching graphite purification strategy template to optimize the gradient density separation and purification process; A Bayesian normalization layer is introduced, and historical graphite purification data is used to simulate the noise quantization of liquid gradient concentration distribution. A graphite purification scenario template library is constructed, including: The historical graphite purification data were sorted out according to physical properties, liquid gradient concentration distribution and purification index to obtain a historical physical properties-liquid gradient concentration distribution-purification index data set. Based on physical characteristics, the historical physical characteristics-liquid gradient concentration distribution-purification index data set is aggregated in the same category to obtain multiple aggregated historical physical characteristics-liquid gradient concentration distribution-purification index data sets. Based on multiple datasets of historical physical properties of polymerization, liquid gradient concentration distribution, and purification indicators, noise fluctuation scales are identified to obtain multiple noise fluctuation scales. A Bayesian normalization layer is introduced to randomly perturb and expand the multiple sets of polymerization history physical properties-liquid gradient concentration distribution-purification index data according to the multiple noise fluctuation scales, thereby obtaining multiple expanded sets of polymerization history physical properties-liquid gradient concentration distribution-purification index data. By centrally sorting out the data of the multiple extended polymerization historical physical properties, liquid gradient concentration distribution and purification index data sets, multiple graphite purification scenario template sets were determined, and the graphite purification scenario template library was constructed.

2. The graphite purification method based on gradient density separation as described in claim 1, characterized in that, Noise fluctuation scales were identified based on multiple datasets of historical polymerization physical properties, liquid gradient concentration distribution, and purification indicators, resulting in multiple noise fluctuation scales, including: By traversing the multiple sets of historical physical properties of polymerization, liquid gradient concentration distribution, and purification index data, liquid gradient concentration distribution fluctuation analysis is performed to determine multiple liquid gradient concentration distribution fluctuation factors. By traversing multiple sets of historical physical properties, liquid gradient concentration distribution, and purification index data, a fluctuation analysis of purification indexes is performed to determine multiple purification index follow-up factors. Based on the multiple liquid gradient concentration distribution fluctuation factors and the multiple purification index follow-up factors, noise fluctuation scales are identified to obtain multiple noise fluctuation scales.

3. The graphite purification method based on gradient density separation as described in claim 1, characterized in that, A Bayesian normalization layer is introduced to randomly perturb and expand the multiple sets of polymerization history physical properties-liquid gradient concentration distribution-purification index data according to the multiple noise fluctuation scales, resulting in multiple expanded sets of polymerization history physical properties-liquid gradient concentration distribution-purification index data, including: The Bayesian normalization layer randomly perturbs the liquid gradient concentration distribution and purification index data in the multiple sets of polymerization history physical properties-liquid gradient concentration distribution-purification index data according to the multiple noise fluctuation scales to obtain multiple initial expanded sets of polymerization history physical properties-liquid gradient concentration distribution-purification index data. The average value of the initial expansion polymerization historical physical properties-liquid gradient concentration distribution-purification index data with an approximation greater than a preset approximation threshold in the multiple initial expansion polymerization historical physical properties-liquid gradient concentration distribution-purification index data sets is processed to obtain the multiple expansion polymerization historical physical properties-liquid gradient concentration distribution-purification index data sets.

4. The graphite purification method based on gradient density separation as described in claim 3, characterized in that, Gaussian noise is generated based on the multiple noise fluctuation scales through the Bayesian normalization layer, and the liquid gradient concentration distribution and purification index data in the multiple polymerization history physical properties-liquid gradient concentration distribution-purification index data sets are adjusted according to the generation results to obtain the multiple initial expanded polymerization history physical properties-liquid gradient concentration distribution-purification index data sets.

5. The graphite purification method based on gradient density separation as described in claim 1, characterized in that, The graphite purification scenario template library is traversed to identify graphite purification strategies, and a graphite purification strategy template library is constructed, including: A pre-built graphite purification strategy recognizer is provided, wherein the graphite purification strategy recognizer includes an input layer, a convolutional layer, and an output layer; The graphite purification strategy recognizer is used to perform strategy recognition on the graphite purification scenario templates in the graphite purification scenario template library to obtain the graphite purification strategy template library.

6. The graphite purification method based on gradient density separation as described in claim 1, characterized in that, Using the physical characteristics of the graphite ore sample and the monitoring results of the liquid gradient concentration distribution during the purification process as indexes, cosine similarity calculations are performed on the graphite purification scene templates in the graphite purification scene template library. The graphite purification scene template corresponding to the maximum calculated value is extracted to obtain the matching graphite purification scene template.

7. The graphite purification method based on gradient density separation as described in claim 1, characterized in that, include: The gradient density separation and purification process is continuously monitored to obtain a sequence of continuous monitoring results of liquid gradient concentration distribution. The deviation sequence is obtained by performing a deviation analysis between the continuous monitoring result sequence of the liquid gradient concentration distribution and the corresponding liquid gradient concentration distribution in the matching graphite purification scenario template. The deviation sequence is analyzed from two dimensions: the trend of change and the central tendency of data, to determine the deviation influence coefficient; The matching graphite purification strategy template is optimized based on the deviation influence coefficient.

8. The graphite purification method based on gradient density separation as described in claim 7, characterized in that, The deviation sequence is analyzed from two dimensions: trend of change and central tendency, to determine the deviation influence coefficient, including: The mean shift algorithm is used to perform clustered data identification on the deviation sequence to obtain the clustered deviation. The deviation sequence is analyzed to obtain the deviation change trend characteristics; The concentration deviation is corrected based on the deviation change trend characteristics to obtain the deviation influence coefficient.

9. A graphite purification system based on gradient density separation, characterized in that, The step of implementing the graphite purification method based on gradient density separation according to any one of claims 1 to 8, wherein the graphite purification system based on gradient density separation comprises: The physical property measurement module is used to screen graphite ore samples and perform ultrasonic cleaning to remove surface impurities. The cleaned graphite ore samples are then subjected to physical property measurements to obtain physical characteristic information of the graphite ore samples. The noise quantization simulation module is used to introduce a Bayesian normalization layer, combine historical graphite purification data to perform noise quantization simulation of liquid gradient concentration distribution, and build a graphite purification scenario template library. The purification strategy identification module is used to traverse the graphite purification scenario template library to identify graphite purification strategies and construct the graphite purification strategy template library. The matching and identification module is used to perform gradient density separation and purification on the graphite ore sample, and combine the physical characteristic information of the graphite ore sample and the monitoring results of liquid gradient concentration distribution during the purification process to match and identify the graphite purification scene template library to obtain a matching graphite purification scene template. The strategy retrieval module is used to search the graphite purification strategy template library based on the matching graphite purification scenario template, and obtain matching graphite purification strategy templates to optimize the gradient density separation and purification process.

Images