Graphite anode life prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a graphite anode life prediction method and device, electronic equipment and a storage medium, and relates to the technical field of rare earth electrolysis. According to the graphite anode life prediction method, physical characteristic parameters and technological parameters of a graphite anode are collected, and the collected physical characteristic parameters and technological parameters are input into an SVM-simulated annealing coupling prediction model after being subjected to abnormal value processing and normalization processing; according to the simulated annealing algorithm, a penalty factor C and a kernel parameter delta of the SVM are optimized through an adaptive temperature updating strategy and a Metropolis criterion, and local optimum is avoided. According to the method, a breakthrough from'artificial experience judgment 'to'mathematical model quantitative prediction' is realized, the problem of low precision of a traditional method is solved, and technical support is provided for graphite anode replacement cycle planning and cost control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rare earth electrolysis, and in particular to a graphite anode service life prediction method and device, an electronic device, and a storage medium. BACKGROUND

[0002] In the rare earth electrolysis industry, the graphite anode is the core functional component of the electrolysis system, and its service life directly determines the production continuity, unit product energy consumption and rare earth product purity. Short service life will lead to frequent replacement, increasing production cost and production interruption risk; inaccurate service life prediction may cause electrolytic cell working condition disorder due to excessive anode wear, affecting product quality stability. From the perspective of wear mechanism, the failure of graphite anode is the result of physical wear (such as electrolyte erosion, structure peeling caused by thermal expansion) and electrochemical corrosion (such as oxidation reaction with electrolyte, air, etc.) under high temperature environment. This process is affected by multiple factors coupling inherent properties of graphite and electrolysis process conditions, and objectively requires life prediction to cover multiple dimensions of key parameters.

[0003] However, the current industry's prediction methods for graphite anode life still remain at the extensive stage, with significant technical limitations, specifically in the following two aspects:

[0004] On the one hand, artificial experience judgment dominates, with strong subjectivity and uncontrollable precision. Existing prediction methods rely heavily on the subjective judgment of front-line operators based on production time, anode appearance (such as surface peeling degree, color change), lacking quantitative standards. The experience accumulation and observation angle of different operators vary greatly, leading to high dispersion of graphite anode life prediction results under the same batch and working conditions. For example, inexperienced personnel may replace the graphite anode prematurely due to misjudgment of "slight corrosion" as "serious failure", causing waste of graphite material; experienced personnel may also delay replacement due to neglecting internal hidden wear (such as increased internal porosity) of the anode, causing electrolytic cell leakage, sudden drop in current efficiency and other production accidents. At the same time, artificial judgment cannot analyze the contribution of each influencing factor (such as graphite density, current density) to the service life, making it difficult to establish a "parameter-life" correlation, and the prediction results cannot guide the customized production and process optimization of graphite anode.

[0005] On the other hand, estimation based on a single parameter has significant limitations and cannot reflect the coupling effect of multiple factors. Some companies attempt to roughly estimate lifespan using a single parameter (such as graphite bulk density or electrolytic current density), for example, believing that "the higher the bulk density, the longer the lifespan" or "the lower the current density, the longer the lifespan." However, this single-dimensional prediction logic completely ignores the synergistic effect between parameters. For instance, a high-bulk-density graphite anode with excessively high resistivity may experience accelerated oxidation due to localized overheating at high current densities, potentially resulting in a shorter lifespan than a low-bulk-density anode with a suitable resistivity. Similarly, while graphite with low air permeability can reduce external oxidation, insufficient flexural strength can lead to thermal stress fracture at high temperatures, also shortening the service life. This "generalizing from limited evidence" prediction method results in a generally large deviation between predicted and actual lifespan values, failing to provide a reliable basis for production planning and spare parts inventory management, ultimately leading to uncontrolled production costs and low production efficiency.

[0006] Furthermore, even in some attempts to introduce simple mathematical tools, implementation has been hampered by the failure to address core technical issues. For example, a few studies have tried to correlate parameters with lifetime using linear regression models, but these models cannot handle the nonlinear relationships in the failure process of graphite anodes (such as the exponential increase in lifetime decay rate after the current density exceeds a critical value). As for algorithms with nonlinear modeling capabilities, such as Support Vector Machines (SVM), the industry has not yet developed parameter optimization schemes. If the penalty factor C (which controls the intensity of penalty for misclassified samples) and kernel parameter δ (which controls the dimension of sample mapping) of the SVM model are set based solely on experience, overfitting (the model only adapts to the training data and fails to predict new samples) or underfitting (the model cannot capture key patterns) is very likely to occur, resulting in prediction accuracy in actual production that is even lower than human experience judgment, ultimately failing to achieve technological implementation.

[0007] In summary, current methods for predicting the lifetime of graphite anodes are no longer able to meet the needs of the rare earth electrolysis industry's transformation towards "precision, low cost, and high efficiency" due to their high subjectivity, incomplete parameter coverage, and poor model adaptability. There is an urgent need to break through existing technological bottlenecks and establish a scientific and quantitative multi-parameter collaborative prediction system. Summary of the Invention

[0008] The purpose of this application is to provide a method for predicting the lifespan of graphite anodes used in rare earth electrolysis based on the physical properties and process parameters of graphite anodes. This method solves the problems of existing prediction methods relying on experience, considering only one factor, and having low accuracy. It enables high-precision prediction of graphite anode lifespan and provides technical support for planning the replacement cycle of graphite anodes and controlling production costs in rare earth electrolysis.

[0009] This application also provides a graphite anode lifetime prediction device, electronic device, and storage medium, which can also achieve high-precision prediction of graphite anode lifetime.

[0010] The first objective of this application is to provide a method for predicting the lifetime of graphite anodes.

[0011] The aforementioned objective of this application is achieved through the following technical solution:

[0012] A method for predicting the lifetime of a graphite anode, the method comprising the following steps:

[0013] S1, Collect the physical property parameters and process parameters of the target graphite anode;

[0014] S2, preprocess the collected physical characteristic parameters and process parameters, wherein the preprocessing includes abnormal data processing and normalization processing;

[0015] S3, input the preprocessed data into the SVM-simulated annealing coupled prediction model to obtain the lifetime prediction result of the target graphite anode, wherein,

[0016] The SVM-simulated annealing coupled prediction model is constructed by optimizing the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm. During the optimization of the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the quality of the current solution in each iteration, and the Metropolis criterion is used to determine whether to accept a new solution.

[0017] Preferably, in step S1,

[0018] The physical property parameters include at least bulk density, resistivity, true density, flexural strength, compressive strength, air permeability, air reactivity, CO2 reactivity, and XRD002 value;

[0019] The process parameters include at least the graphite current density.

[0020] Preferably, in step S2,

[0021] The abnormal data processing includes: using the 3σ criterion to identify abnormal data, and filling the identified abnormal data with the mean of adjacent samples;

[0022] The normalization process includes: using the min-max normalization formula to unify the dimensions of each parameter.

[0023] Preferably, in step S3, the model of the adaptive temperature update strategy of the simulated annealing algorithm is as follows:

[0024]

[0025] In the formula, For the first Temperature of the cycle iteration This is the difference between the current solution and the optimal solution. This represents the number of iterations.

[0026] Preferably, in step S3, the formula for calculating the acceptance probability P of the new solution of the Metropolis criterion is:

[0027]

[0028] In the formula, This represents the change in prediction error between the new solution and the current solution. This is the current temperature.

[0029] Preferably, in step S3, the kernel function of the SVM model is as follows:

[0030]

[0031] In the formula, Represents the kernel function value. For kernel parameters, , This is the input sample.

[0032] Preferably, in step S3, the SVM model training process employs five-fold cross-validation, using the root mean square error as the fitness function, as shown in the formula:

[0033]

[0034] In the formula, This represents the root mean square error. For the sample size, This is the actual lifespan value. To predict lifetime values.

[0035] The second objective of this application is to provide a graphite anode lifetime prediction device.

[0036] The second objective of this application is achieved through the following technical solution:

[0037] A graphite anode lifetime prediction device, the device comprising:

[0038] The data acquisition module is used to collect the physical property parameters and process parameters of the target graphite anode;

[0039] The data preprocessing module is used to preprocess the collected physical characteristic parameters and process parameters, wherein the preprocessing includes abnormal data processing and normalization processing.

[0040] The lifetime prediction module is used to input the preprocessed data into the SVM-simulated annealing coupled prediction model to obtain the lifetime prediction result of the target graphite anode, wherein...

[0041] The SVM-simulated annealing coupled prediction model is constructed by optimizing the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm. During the optimization of the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the quality of the current solution in each iteration, and the Metropolis criterion is used to determine whether to accept a new solution.

[0042] Preferably, the physical property parameters include at least bulk density, resistivity, true density, flexural strength, compressive strength, air permeability, air reactivity, CO2 reactivity, and XRD002 value;

[0043] The process parameters include at least the graphite current density.

[0044] Preferably,

[0045] The abnormal data processing includes: using the 3σ criterion to identify abnormal data, and filling the identified abnormal data with the mean of adjacent samples;

[0046] The normalization process includes: using the min-max normalization formula to unify the dimensions of each parameter.

[0047] Preferably, the adaptive temperature update strategy model of the simulated annealing algorithm is as follows:

[0048]

[0049] In the formula, For the first Temperature of the cycle iteration This is the difference between the current solution and the optimal solution. This represents the number of iterations.

[0050] Preferably, the formula for calculating the acceptance probability P of the new solution of the Metropolis criterion is:

[0051]

[0052] In the formula, This represents the change in prediction error between the new solution and the current solution. This is the current temperature.

[0053] Preferably, the kernel function of the SVM model is as follows:

[0054]

[0055] In the formula, Represents the kernel function value. For kernel parameters, , This is the input sample.

[0056] Preferably, the SVM model employs five-fold cross-validation during training, using the root mean square error as the fitness function, as shown in the formula:

[0057]

[0058] In the formula, This represents the root mean square error. For the sample size, This is the actual lifespan value. To predict lifetime values.

[0059] The third objective of this application is to provide an electronic device.

[0060] The aforementioned objective three of this application is achieved through the following technical solution:

[0061] An electronic device, comprising:

[0062] The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the graphite anode lifetime prediction method described in any of the first objectives of this application.

[0063] The fourth objective of this application is to provide a computer-readable storage medium.

[0064] The fourth objective of this application is achieved through the following technical solution:

[0065] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the graphite anode lifetime prediction method described in any one of the first objectives of this application.

[0066] In summary, this application discloses a method, apparatus, electronic device, and storage medium for predicting the lifetime of graphite anodes. By collecting the physical and process parameters of the graphite anode, and after preprocessing, inputting them into an SVM-simulated annealing coupled prediction model, the model employs a simulated annealing algorithm to optimize the penalty factor C and kernel parameter δ of the SVM model. It avoids local optima through an adaptive temperature update strategy and the Metropolis criterion, effectively improving prediction accuracy. This solves the problems of existing prediction methods relying on experience, considering only a single factor, and having low accuracy, achieving high-precision prediction of graphite anode lifetime and providing technical support for the replacement cycle planning and production cost control of graphite anodes in rare earth electrolysis. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a flowchart illustrating a graphite anode lifetime prediction method according to an embodiment of this application.

[0069] Figure 2 The graph shows a comparison between the predicted values ​​of the original model (SVM model) and the optimized model (SVM-simulated annealing coupled prediction model) of a graphite anode lifetime prediction method in this application embodiment and the actual values.

[0070] Figure 3 This is a schematic diagram of the structure of a graphite anode lifetime prediction device according to an embodiment of this application;

[0071] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0072] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0073] In the embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described below are merely illustrative. For example, the division of units and modules is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or modules can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, and can be electrical, mechanical, or other forms.

[0074] In addition, each functional unit in the various embodiments of this application can be integrated into a single processor, or each unit can be a separate device, or two or more units can be integrated into a single device; each functional unit in the various embodiments of this application can be implemented in hardware or in the form of hardware plus software functional units.

[0075] Those skilled in the art will understand that all or part of the steps of the following method embodiments can be implemented by program instructions and related hardware. The aforementioned program instructions can be stored in a computer-readable storage medium. When the program instructions are executed, they perform the steps of the following method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0076] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "a plurality of" or "several" means two or more, unless otherwise explicitly specified.

[0077] Existing methods for predicting the lifetime of graphite anodes rely entirely on manual experience or single-parameter estimation, and have never employed any mathematical models, computer algorithms, or multi-parameter collaborative analysis to predict the lifetime of graphite anodes. While Support Vector Machines (SVMs) have unique advantages in small-sample, multi-parameter modeling, the selection of their model parameters (penalty factor C, kernel parameter δ) significantly affects prediction accuracy.

[0078] Therefore, the purpose of this application is to provide a method for predicting the lifetime of graphite anodes for rare earth electrolysis based on the physical properties and process parameters of graphite anodes, so as to achieve a breakthrough from "human experience judgment" to "mathematical model quantitative prediction", fill the technical gap in the field of predicting the lifetime of graphite anodes using computer models, and significantly improve the prediction accuracy of graphite anode lifetime in rare earth electrolysis.

[0079] like Figure 1 As shown in the figure, this application provides a method for predicting the lifetime of a graphite anode, which may include the following steps:

[0080] S1, Collect the physical property parameters and process parameters of the target graphite anode;

[0081] To quantitatively predict the lifespan of the target graphite anode using a mathematical model, it is first necessary to collect relevant parameters of the target graphite anode. To ensure that the collected parameters more accurately reflect the lifespan status of the target graphite anode, this embodiment collects the physical characteristic parameters and process parameters of the target graphite anode. These physical characteristic parameters and process parameters, from the two dimensions of the graphite anode's inherent properties and operating conditions, jointly determine the wear rate and lifespan of the target graphite anode under electrolytic conditions.

[0082] S2, preprocessing the collected physical characteristic parameters and process parameters, including abnormal data processing and normalization processing;

[0083] After collecting the physical and process parameters of the target graphite anode, the collected physical and process parameters are further preprocessed, such as abnormal data processing and normalization processing, to eliminate data noise and dimensional differences, and provide high-quality input for the mathematical model.

[0084] S3, the preprocessed data is input into the SVM-simulated annealing coupled prediction model to obtain the lifetime prediction results of the target graphite anode, where,

[0085] The SVM-simulated annealing coupled prediction model is constructed by optimizing the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm. In the process of optimizing the penalty factor C and kernel parameter δ of the SVM model through the simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the quality of the current solution in each iteration, and the Metropolis criterion is used to determine whether to accept a new solution.

[0086] Finally, the preprocessed data is input into the pre-constructed SVM-simulated annealing coupled prediction model, which outputs the corresponding results to obtain the lifetime prediction results of the target graphite anode. The SVM model has the ability to handle small samples and complex relationships between multiple parameters, and can establish a mapping relationship between parameters and graphite anode lifetime. By combining the simulated annealing algorithm to optimize the penalty factor C and kernel parameter δ of the SVM model, a suitable SVM-simulated annealing coupled prediction model for predicting graphite anode lifetime is obtained.

[0087] In the process of optimizing the penalty factor C and kernel parameter δ of the SVM model through simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the difference between the current solution and the optimal solution in each iteration and the number of iterations. At the same time, combined with the Metropolis criterion, a certain number of poor solutions are allowed during the optimization process to avoid getting trapped in local optima. This accurately optimizes the penalty factor C and kernel parameter δ of the SVM model and ensures that the obtained SVM-simulated annealing coupled prediction model has the best prediction performance.

[0088] In summary, the graphite anode lifetime prediction method in this embodiment first collects the physical and process parameters of the target graphite anode. Then, it performs preprocessing on the collected physical and process parameters, including anomaly processing and normalization. Finally, it inputs the preprocessed data into an SVM-simulated annealing coupled prediction model to obtain the lifetime prediction result of the target graphite anode. This prediction method integrates multiple influencing factors, overcoming the limitations of traditional methods that rely on single parameters or empirical judgments, and achieving multi-parameter collaborative analysis. The optimization mechanism of the simulated annealing algorithm effectively solves the problem of SVM model parameter selection, improves model prediction accuracy, and provides a scientific basis for graphite anode replacement cycle planning. This helps reduce cost waste caused by premature replacement or production failures caused by delayed replacement, thereby reducing production costs and improving production efficiency.

[0089] In one embodiment, in step S1, the physical property parameters include at least bulk density, resistivity, true density, flexural strength, compressive strength, air permeability, air reactivity, CO2 reactivity, and XRD002 value; the process parameters include at least graphite current density.

[0090] In this embodiment, among the parameters collected from the target graphite anode, several physical characteristic parameters that are highly correlated with the lifetime of the target graphite anode, such as bulk density, resistivity, true density, flexural strength, compressive strength, air permeability, air reactivity, CO2 reactivity, and XRD002 value, and the process parameter of graphite current density, which is also highly correlated with the lifetime of the target graphite anode, are selected to effectively ensure the accuracy of subsequent lifetime prediction results.

[0091] Specifically, among the above-mentioned physical properties of graphite anodes, the bulk density can be determined by the water displacement method, the resistivity by the four-probe method, the true density by the gas displacement method (helium method), the flexural strength by the three-point bending method, the compressive strength by the axial compression method, the air permeability by the steady-state method, the air reactivity by the thermogravimetric analysis method, and the XRD002 value by the X-ray diffraction method; among the process parameters, the graphite current density is determined by the current-area method.

[0092] The specific procedures for measuring the above parameters are existing technologies and will not be elaborated here. It should be noted that the collection of these parameters all followed relevant industry testing standards to ensure the accuracy and repeatability of the data, providing reliable input for subsequent model training.

[0093] In other embodiments, the collected physical property parameters may also include the porosity and ash content of graphite. Porosity is collected using mercury porosimetry, with a measurement range of 0.003-1000 μm. Ash content is determined using a high-temperature calcination method (calcination at 900℃±10℃ to constant weight). The collected process parameters also include electrolysis temperature, electrolyte component concentration, and electrolysis time. The electrolysis temperature is collected with an accuracy of ±1℃, and the electrolyte component concentration is determined by X-ray fluorescence spectrometry with a measurement error not exceeding 0.1%. By collecting more parameters as input to the prediction model, the accuracy of the prediction results can be improved to some extent.

[0094] In one embodiment, in step S2,

[0095] Outlier processing includes: using the 3σ criterion to identify outlier data, and filling the identified outlier data with the mean of adjacent samples;

[0096] Normalization includes: using the min-max standardization formula to unify the dimensions of each parameter.

[0097] In outlier handling, the 3σ criterion, based on the normal distribution characteristics of the data, considers data falling outside the mean range μ±3σ as outliers. This criterion can accurately identify abnormal data generated during the data collection process due to equipment errors, human operational mistakes, etc. Filling outliers with the mean of adjacent samples can preserve the overall trend and local correlation of the data to the greatest extent, avoiding data loss caused by directly deleting outliers. This outlier handling method can effectively improve data quality and reduce the negative impact of noise on model prediction results.

[0098] In the normalization process, min-max normalization maps all parameter values ​​to the [0,1] interval, eliminating differences in the units (e.g., bulk density is in g / cm³, resistivity is in μΩ·m) and magnitudes (e.g., air permeability may be 10⁻¹).7 To mitigate the interference of current density (on the order of hundreds) on SVM model training, this normalization method ensures that each parameter has equal weight during training, guaranteeing that the model learns the relationship between each parameter and the graphite anode lifetime fairly. This method effectively unifies the parameter scale, ensuring the stability and consistency of model training, preventing bias towards parameters with larger magnitudes due to differences in parameter magnitudes, improving the model's efficiency in utilizing parameter information, and ultimately enhancing the model's prediction accuracy.

[0099] In one embodiment, in step S3, the model of the adaptive temperature update strategy of the simulated annealing algorithm is as follows:

[0100]

[0101] In the formula, For the first Temperature of the cycle iteration This is the difference between the current solution and the optimal solution. This represents the number of iterations.

[0102] In the optimization process of the simulated annealing algorithm, temperature is a key parameter controlling the algorithm's search behavior. In the adaptive temperature update strategy, when... When the difference between the current solution and the optimal solution is large, it indicates a significant gap between them. In this case, the exponential value is small, the temperature decreases slowly, and the algorithm can perform a wider search at a higher temperature, increasing the probability of finding a better solution. When the value is small, it indicates that the current solution is close to the optimal solution. As the exponential value increases, the temperature decreases more rapidly, and the algorithm can quickly converge to the vicinity of the optimal solution, reducing unnecessary search steps. Simultaneously, with the number of iterations... As the temperature increases, the overall temperature decreases, which is consistent with the gradual decrease in temperature during the simulated physical annealing process, ensuring that the algorithm has strong global search capabilities in the early stages and strong local search capabilities in the later stages.

[0103] Compared to traditional simulated annealing algorithms with a fixed cooling rate, the adaptive temperature update strategy can dynamically adjust the cooling rhythm according to the actual situation of the optimization process, balance the global search capability and local search capability of the algorithm, and significantly shorten the optimization time while ensuring that the global optimal solution is found, thereby improving the efficiency of SVM model parameter optimization and providing a guarantee for the rapid deployment of the model in practical applications.

[0104] In one embodiment, in step S3, the formula for calculating the acceptance probability P of the new solution to the Metropolis criterion is:

[0105]

[0106] In the formula, This represents the change in prediction error between the new solution and the current solution. This is the current temperature.

[0107] During the iterative search for new solutions using the simulated annealing algorithm, when the energy difference between the new solutions (i.e., the change in the prediction error between the new solutions and the current solutions) is reached... When the error is less than 0, it indicates that the prediction error of the new solution is smaller than that of the current solution, and the new solution is better. In this case, according to the Metropolis criterion, the acceptance probability P=1, and the new solution is directly accepted, achieving iterative optimization of the solution; when... When the error is greater than or equal to 0, the prediction error of the new solution is greater than that of the current solution. The new solution is relatively worse, but not absolutely inferior. In this case, the acceptance probability is... Furthermore, the P value decreases as the temperature T decreases. In the early stages of the algorithm's search, when the temperature T is high, the P value is large, and the algorithm has a higher probability of accepting a poor solution, allowing it to escape the current local optimum and perform a global search. As the temperature T gradually decreases, the P value gradually decreases, the probability of the algorithm accepting a poor solution decreases, and it gradually converges to the vicinity of the global optimum, avoiding stagnation at a local optimum.

[0108] The Metropolis criterion effectively avoids the simulated annealing algorithm from getting trapped in local optima, ensuring that the algorithm can search for the global optimum of the SVM model parameters. This gives the SVM model the best predictive performance, reduces the problem of decreased prediction accuracy caused by parameters getting trapped in local optima, and further improves the accuracy of graphite anode lifetime prediction.

[0109] In one embodiment, in step S3, the kernel function of the SVM model is as follows:

[0110]

[0111] In the formula, Represents the kernel function value. For kernel parameters, , This is the input sample.

[0112] The SVM model maps input samples from the original low-dimensional space to a high-dimensional feature space through a kernel function, constructing a linear classification surface or regression function in the high-dimensional space to address the nonlinearity of the data in the original space. In this embodiment, the SVM model uses the aforementioned radial basis function as its kernel function. Radial basis functions possess good locality and nonlinear mapping capabilities, and their function values ​​depend only on the samples. and The Euclidean distance between them, through kernel parameters The radial range of the control function. When When the radius is small, the radial basis function has a wider range of influence, resulting in better model smoothness; when... When the value is large, the effective range of the radial basis function is narrower, and the model becomes more sensitive to local samples. In the relationship between graphite anode parameters and lifetime, the influence of each parameter on lifetime exhibits complex nonlinear characteristics. The radial basis function can effectively capture this nonlinear relationship, mapping multi-dimensional parameter data to a suitable high-dimensional space, enabling the SVM model to accurately learn the mapping law between parameters and lifetime.

[0113] The introduction of radial basis function (RBF) kernels enables SVM models to handle the nonlinear relationship between graphite anode parameters and lifetime. Compared with linear or polynomial kernels, RBF kernels have advantages such as fewer parameters, lower computational cost, and stronger generalization ability. They can reduce model complexity and training difficulty while ensuring model prediction accuracy, thereby improving the model's practicality and reliability.

[0114] In one embodiment, in step S3, five-fold cross-validation is used during the training of the SVM model, with the root mean square error as the fitness function, as shown in the formula:

[0115]

[0116] In the formula, This represents the root mean square error. For the sample size, This is the actual lifespan value. To predict lifetime values.

[0117] Five-fold cross-validation randomly divides the training set into five equal parts. Each time, four parts are selected as the training subset for model training, and the remaining part is used as the validation subset to evaluate model performance. This process is repeated five times, ensuring that each data set can participate in model evaluation as a validation subset. The average of the five evaluation results is then used as the final performance metric for the model. This validation method fully utilizes limited training data, comprehensively evaluates the model's generalization ability across different data subsets, and avoids performance evaluation bias caused by improper data partitioning. Root Mean Square Error (RMSE) quantifies the model's prediction error by calculating the square root of the average of the squares of the differences between the actual and predicted lifetimes. A smaller RMSE value indicates higher prediction accuracy. Using RMSE as a fitness function provides a clear target direction for the simulated annealing algorithm to optimize SVM model parameters, enabling the algorithm to search for optimal parameters in the direction of reducing prediction error.

[0118] Five-fold cross-validation effectively improves the accuracy and reliability of model performance evaluation and avoids overfitting (i.e., the model performs well on the training set, but the prediction accuracy drops significantly on new data). Using RMSE as the fitness function gives the parameter optimization process of the simulated annealing algorithm a clear objective, ensuring that the optimized SVM model has the lowest prediction error, further improving the model's prediction accuracy and practicality.

[0119] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of the construction and training process of the SVM-simulated annealing coupled prediction model in the graphite anode lifetime prediction method of this application is provided in conjunction with specific embodiments.

[0120] 1. Data Acquisition and Preprocessing

[0121] Fifty-two graphite anode samples were selected from a rare earth smelter, and the parameters collected included:

[0122] Physical property parameters (9 items):

[0123] Bulk density (1.672-1.721 g / cm³), resistivity (6-23.55 μΩ・m), true density (2.169-2.241 g / cm³), flexural strength (9.05-18.82 MPa), compressive strength (28.95-35.9 MPa), air permeability (0.392-1.02×10⁻¹) 7 m²), air reactivity (0.9043-0.9758 mm / h), CO2 reactivity (0.9043-0.9758 mm / h), XRD002 value (0.33657-0.33958);

[0124] Process parameters (1 item):

[0125] Graphite current density (593-827 A / cm²);

[0126] Output metrics:

[0127] The actual service life of graphite anodes is 37-48 days.

[0128] 2. Outlier Handling

[0129] Use the 3σ criterion to identify outlier data:

[0130] Calculate the mean μ and standard deviation σ of each parameter;

[0131] Judgment exceeds The data within the range is outlier;

[0132] Outliers are filled with the mean of adjacent samples.

[0133] For example: if the volume density of a sample is 1.9 g / cm³, exceeding... The range was filled with the average volume density of the samples before and after it (1.71+1.73) / 2=1.72g / cm³.

[0134] 3. Normalization processing

[0135] The min-max standardization formula is used to unify the dimensions:

[0136]

[0137] in, These are the original parameter values. , These are the minimum and maximum values ​​of the parameter, respectively. This is the normalized value.

[0138] Example: Bulk density x = 1.70 g / cm³, , Then y = (1.70 - 1.672) / (1.721 - 1.672) = 0.571.

[0139] Inverse normalization: If a parameter is normalized to y=0.6, , Then the original value x = 0.392 + (1.02 - 0.392) × 0.6 = 0.769.

[0140] 4. Construction of SVM-Simulated Annealing Coupled Model

[0141] SVM model initialization:

[0142] Kernel function: Radial basis function is used.

[0143]

[0144] Where δ is the kernel parameter, and xi and xj are the input samples.

[0145] Parameter to be optimized: Penalty factor nuclear parameters .

[0146] Simulated annealing algorithm optimization parameters:

[0147] The simulated annealing algorithm is used to search for the optimal parameters C and δ, with the following specific settings:

[0148] Initial parameters: initial temperature T0=100, maximum number of iterations 50, number of iterations per temperature 20;

[0149] Adaptive temperature update strategy:

[0150]

[0151] In the formula, For the first Temperature of the cycle iteration This is the difference between the current solution and the optimal solution. This represents the number of iterations.

[0152] Metropolis Criterion: The probability P of accepting a new solution is:

[0153]

[0154] in, The change in prediction error (energy difference) between the new solution and the current solution is represented by T, where T is the current temperature.

[0155] Optimize the process:

[0156] Randomly generate initial parameters and Calculate the initial fitness (RMSE of five-fold cross-validation).

[0157] Generate a new solution: Add random perturbations to the current parameters, constraining them within the parameter range;

[0158] Calculate the fitness of the new solution and determine whether to accept the new solution using the Metropolis criterion;

[0159] Update the temperature according to the adaptive temperature formula, iterate until the maximum number of iterations, and output the optimal parameters. and .

[0160] 5. Model Training and Prediction

[0161] Model training:

[0162] Cross-validation: Five-fold cross-validation is used, dividing the training set into 5 parts, using 4 parts for training and 1 part for validation in turn, repeating 5 times;

[0163] Fitness function: The root mean square error (RMSE) is used as the fitness function.

[0164]

[0165] Where N is the sample size. For actual lifespan, To predict lifespan.

[0166] Convergence Criterion: Iteration stops when the RMSE of the training set converges to less than 3.5 days. In this embodiment, the optimal parameters are: , .

[0167] predict:

[0168] The preprocessed test set data is input into the trained model, which outputs the predicted lifetime value of the graphite anode:

[0169]

[0170] Result evaluation:

[0171] To evaluate the prediction results, the root mean square error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) are introduced to evaluate the model's prediction results. The expressions are as follows:

[0172]

[0173]

[0174]

[0175] Evaluation of the optimized SVM model:

[0176] Training set: MSE=0.0359, RMSE=0.1895, MAE=0.1731

[0177] Test set: MSE=0.9275, RMSE=0.9630, MAE=0.5233

[0178] The original model's MSE on the test set was 3.9477, which was reduced by 76.51% after optimization.

[0179] pass Figure 2 It can be seen that the SVM-simulated annealing coupled prediction model obtained by the optimization of the SVM algorithm in this application has a better fit between the lifetime prediction value and the actual lifetime value than the original SVM model. This indicates that the SVM-simulated annealing coupled prediction model obtained by optimizing the SVM model by the annealing algorithm in this application can effectively improve the accuracy of graphite anode lifetime prediction.

[0180] like Figure 3 As shown, this application provides a graphite anode lifetime prediction device, which may include:

[0181] The data acquisition module 201 is used to acquire the physical property parameters and process parameters of the target graphite anode;

[0182] The data preprocessing module 202 is used to preprocess the collected physical characteristic parameters and process parameters, wherein the preprocessing includes abnormal data processing and normalization processing;

[0183] The lifetime prediction module 203 is used to input the preprocessed data into the SVM-simulated annealing coupled prediction model to obtain the lifetime prediction results of the target graphite anode.

[0184] The SVM-simulated annealing coupled prediction model is constructed by optimizing the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm. In the process of optimizing the penalty factor C and kernel parameter δ of the SVM model through the simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the quality of the current solution in each iteration, and the Metropolis criterion is used to determine whether to accept a new solution.

[0185] In one embodiment, the physical property parameters include at least bulk density, resistivity, true density, flexural strength, compressive strength, air permeability, air reactivity, CO2 reactivity, and XRD002 value.

[0186] The process parameters include at least the graphite current density.

[0187] In one embodiment,

[0188] Outlier processing includes: using the 3σ criterion to identify outlier data, and filling the identified outlier data with the mean of adjacent samples;

[0189] Normalization includes: using the min-max standardization formula to unify the dimensions of each parameter.

[0190] In one embodiment, the adaptive temperature update strategy of the simulated annealing algorithm is modeled as follows:

[0191]

[0192] In the formula, For the first Temperature of the cycle iteration This is the difference between the current solution and the optimal solution. This represents the number of iterations.

[0193] In one embodiment, the probability P of accepting a new solution according to the Metropolis criterion is calculated as follows:

[0194]

[0195] In the formula, This represents the change in prediction error between the new solution and the current solution. This is the current temperature.

[0196] In one embodiment, the kernel function of the SVM model is as follows:

[0197]

[0198] In the formula, Represents the kernel function value. For kernel parameters, , This is the input sample.

[0199] In one embodiment, the SVM model uses five-fold cross-validation during training, with the root mean square error as the fitness function, as shown in the formula:

[0200]

[0201] In the formula, This represents the root mean square error. For the sample size, This is the actual lifespan value. To predict lifetime values.

[0202] It should be noted that the graphite anode lifetime prediction device in the above embodiments has the same working principle and technical effect as the graphite anode lifetime prediction method in the above embodiments, and will not be repeated here.

[0203] like Figure 4 As shown, this application provides an electronic device 3, which includes a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and executable on the processor 302. When the processor 302 executes the computer program 303, it implements the steps of the graphite anode lifetime prediction method as described in the above method embodiment of this application.

[0204] Specifically, the electronic device 3 can be an intelligent device with memory and processor, such as an industrial control computer, PC, or smart mobile terminal, or a computer component with memory and processor, such as a CPU or GPU. In this embodiment, the electronic device 3 is an industrial control computer.

[0205] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the graphite anode lifetime prediction method as described in the above-described method embodiments of this application.

[0206] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0207] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0208] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly using hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0209] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the lifetime of a graphite anode, characterized in that, The method includes the following steps: S1, Collect the physical property parameters and process parameters of the target graphite anode; S2, preprocess the collected physical characteristic parameters and process parameters, wherein the preprocessing includes abnormal data processing and normalization processing; S3, input the preprocessed data into the SVM-simulated annealing coupled prediction model to obtain the lifetime prediction result of the target graphite anode, wherein, The SVM-simulated annealing coupled prediction model is constructed by optimizing the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm. During the optimization of the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the quality of the current solution in each iteration, and the Metropolis criterion is used to determine whether to accept a new solution.

2. The graphite anode lifetime prediction method according to claim 1, characterized in that, In step S1, The physical property parameters include at least bulk density, resistivity, true density, flexural strength, compressive strength, air permeability, air reactivity, CO2 reactivity, and XRD002 value; The process parameters include at least the graphite current density.

3. The graphite anode lifetime prediction method according to claim 1, characterized in that, In step S2, The abnormal data processing includes: using the 3σ criterion to identify abnormal data, and filling the identified abnormal data with the mean of adjacent samples; The normalization process includes: using the min-max normalization formula to unify the dimensions of each parameter.

4. The graphite anode lifetime prediction method according to claim 1, characterized in that, In step S3, the model of the adaptive temperature update strategy of the simulated annealing algorithm is as follows: In the formula, For the first Temperature of the cycle iteration This is the difference between the current solution and the optimal solution. This represents the number of iterations.

5. The graphite anode lifetime prediction method according to claim 1, characterized in that, In step S3, the formula for calculating the acceptance probability P of the new solution to the Metropolis criterion is as follows: In the formula, This represents the change in prediction error between the new solution and the current solution. This is the current temperature.

6. The graphite anode lifetime prediction method according to claim 1, characterized in that, In step S3, the kernel function of the SVM model is as follows: In the formula, Represents the kernel function value. For kernel parameters, , This is the input sample.

7. The graphite anode lifetime prediction method according to claim 1, characterized in that, In step S3, the SVM model training process employs five-fold cross-validation, using the root mean square error as the fitness function, as shown in the formula: In the formula, This represents the root mean square error. For the sample size, This is the actual lifespan value. To predict lifetime values.

8. A graphite anode lifetime prediction device, characterized in that, The device includes: The data acquisition module is used to collect the physical property parameters and process parameters of the target graphite anode; The data preprocessing module is used to preprocess the collected physical characteristic parameters and process parameters, wherein the preprocessing includes abnormal data processing and normalization processing. The lifetime prediction module is used to input the preprocessed data into the SVM-simulated annealing coupled prediction model to obtain the lifetime prediction result of the target graphite anode, wherein... The SVM-simulated annealing coupled prediction model is constructed by optimizing the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm. During the optimization of the penalty factor C and kernel parameter δ of the SVM model using the simulated annealing algorithm, an adaptive temperature update strategy is adopted to dynamically adjust the cooling rate according to the quality of the current solution in each iteration, and the Metropolis criterion is used to determine whether to accept a new solution.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the graphite anode lifetime prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the graphite anode lifetime prediction method as described in any one of claims 1-7.

Images