Preparation method of low-power-consumption anode carbon block and anode carbon block finished product

Abstract

The application provides a low-power-consumption anode carbon block preparation method and an anode carbon block finished product, relates to the technical field of industrial data mining, and realizes deep quantification and dynamic adaptation of the energy efficiency conversion potential of the whole chain of anode carbon block preparation by constructing a collaborative evolution model based on finite element analysis logic and big data mining; the robustness to production fluctuations is significantly enhanced by performing time sequence alignment and linkage compensation based on batch identification; the variation coefficient of the coking interface in the finished anode is stably below 0.08, and the uniformity improvement rate of the volume density distribution is higher than 10%, so that the problems of product consistency caused by raw material fluctuations are solved while the anode pressure drop and the electrolysis power consumption are effectively reduced, and solid support is provided for the low-carbonization and intelligent preparation of the aluminum electrolysis industry.

Description

Technical Field

[0001] This invention relates to the field of industrial data mining technology, specifically to a method for preparing low-power anode carbon blocks and the finished anode carbon block product. Background Technology

[0002] Reducing electricity consumption in aluminum electrolysis production through carbon anode blocks is a systematic project, primarily achieved by improving the quality of the anode itself and optimizing its performance in the electrolytic cell. As the "heart" of the electrolytic cell, the anode's performance directly affects electrolysis efficiency.

[0003] The existing technology, with publication number CN120525277A, entitled "A Raw Material Blending Process for Anode Carbon Block Production," includes the following steps: S1, multi-stage crushing and innovative surface modification treatment of aggregates such as calcined petroleum coke and pitch coke, and precise pre-melting and modification treatment of binders such as coal tar pitch; S2, constructing a comprehensive raw material characteristic database, clarifying the target performance indicators of anode carbon blocks, and using cutting-edge intelligent optimization algorithms to accurately calculate the optimal raw material ratio; by optimizing the product operation process and interface design, complex operations are simplified to one-click or several-step processes, reducing the learning and usage difficulty for users, improving operational efficiency, and enabling both novice and experienced users to quickly get started; furthermore, by introducing a personalized customization module, users can freely adjust product functions, interface, and parameters according to their preferences, habits, and scenarios to create a customized usage environment, enhancing user satisfaction and loyalty.

[0004] However, in existing technologies, the mixing, molding, and roasting processes in traditional processes have long been in a state of isolated control, resulting in serious "information gaps" and feedback lags. This means that microscopic data characterizing the aggregate wetting quality during the mixing stage cannot be transmitted in real time to the subsequent roasting stage, making it difficult to dynamically adapt the roasting temperature curve to the actual physical evolution of the raw blocks, ultimately causing drastic fluctuations in the anode resistivity of the finished product. At the same time, the industry generally lacks a quantitative model for full-chain energy efficiency conversion. Existing technologies only focus on achieving single static parameters such as temperature and density, failing to quantitatively describe the energy transfer efficiency between "initial mechanical power consumption input" and "final dense structure formation," making it difficult to monitor and accurately analyze the hidden structural defects inside the carbon blocks in real time.

[0005] The existing production system lacks effective self-healing capabilities to handle raw material batch fluctuations and process disturbances. When random fluctuations occur in coke particle size distribution or binder properties, the control logic cannot automatically compensate through parameter linkage in subsequent processes. This lack of cross-process closed-loop correction capability makes it easy for non-conforming batches to occur when fluctuations occur, severely restricting the performance stability and energy efficiency of anode carbon blocks throughout their entire production lifecycle. Summary of the Invention

[0006] The purpose of this invention is to provide a method for preparing low-power anode carbon blocks and a finished anode carbon block, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for preparing low-power anode carbon blocks, comprising the following steps:

[0009] S1. Extract the first feature vector characterizing the aggregate binder wetting characteristics of the anode carbon block to be prepared during the kneading process, and the second feature vector characterizing the thermodynamic characteristics of the anode carbon block during the calcination process.

[0010] S2. Construct the multidimensional microstructure evolution potential energy matrix of the anode carbon block, use the first eigenvector as the spatial dimension constraint operator and the second eigenvector as the time dimension driving operator, and establish a microstructure evolution simulation model that characterizes the dynamic evolution of the anode carbon block from raw block to finished product.

[0011] S3. Based on the microstructure evolution simulation model, perform spatiotemporal scale alignment operation based on batch identifier to generate energy structure co-evolution signal that quantifies the energy efficiency conversion loss of the entire chain;

[0012] S4. Analyze the energy structure co-evolution signal, calculate the linkage compensation command for the current batch, and adjust the raw material ratio parameters and calcination temperature gradient curve of the next batch accordingly to achieve closed-loop preparation of anode carbon blocks with cross-process performance feedback compensation.

[0013] Furthermore, in S1, the process of extracting the first feature vector includes:

[0014] Acquire the current timing data of the kneading drive motor during the kneading process; perform a sliding window Fourier transform on the current timing data to calculate the power consumption spectral density statistical moments under different time windows; and calculate the porous medium wave impedance evolution rate, which characterizes the energy transfer efficiency inside the anode carbon block material, based on the power consumption spectral density statistical moments.

[0015] Using the power spectral entropy of the current time series data in the frequency domain, the microscopic wetting degree at the interface between the aggregate and the binder during the mixing process of the anode carbon block is quantified, and the microscopic wetting degree is used as the weight gain of the wave impedance evolution rate of the porous medium to construct the first feature vector.

[0016] Furthermore, in S1, the process of extracting the second feature vector includes acquiring thermodynamic multi-source sensing data of the anode carbon block in the roasting process, wherein the thermodynamic multi-source sensing data includes furnace temperature gradient time-series signal and flue gas flow channel heat flux density signal.

[0017] The thermodynamic energy efficiency distribution entropy of the thermodynamic multi-source sensing data is calculated to characterize the energy dissipation characteristics of the anode carbon block during the thermodynamic evolution process; the thermal strain energy density variational characteristics of the anode carbon block at different calcination temperature ranges are extracted, and the second feature vector is constructed by combining the thermodynamic energy efficiency distribution entropy.

[0018] Further, in step S2, the spatial dimension constraint operator specifically executes the following logic: using the first eigenvector as the boundary condition mapping parameter of the potential energy matrix to limit the upper limit of energy transfer at different grid points; the time dimension driving operator specifically executes the following logic: using the second eigenvector as the input weight of the time step sequence to dynamically correct the evolution rate of nonlinear phase transition points in the matrix.

[0019] The process of constructing the multidimensional microstructure evolution potential matrix includes: retaining the original information of the first eigenvector and the second eigenvector through the identity mapping branch, extracting the evolution perturbation of the carbon block structure through the nonlinear mapping branch, and performing element-level addition operation on the original information and the evolution perturbation before performing tensor dimension reshaping.

[0020] Furthermore, the calculation logic of the multidimensional microstructure evolution potential energy matrix includes: establishing the energy distribution mapping relationship of the anode carbon block material during the transformation process from a loose particulate state to a continuous medium state; identifying the potential energy evolution poles of the anode carbon block feature vector at different preparation stages in real time, and using them as input features for nonlinear causal mapping in the simulation model.

[0021] Furthermore, the process of performing spatiotemporal scale alignment in step S3 includes: associating the first feature vector with the corresponding production batch identifier and temporarily storing it in the virtual batch shadow memory; when the batch identifier enters the roasting process, retrieving the corresponding first feature vector from the memory to achieve alignment of the kneading feature with the sampling frequency of minutes and the roasting feature with the sampling frequency of days on the simulation time axis.

[0022] Furthermore, the process of generating the energy structure co-evolution signal includes: calculating the cosine similarity between the aligned first feature vector and the second feature vector in the simulation model; extracting the co-gain deviation value and converting it into the cross-process performance coupling correlation degree as a quantitative index of the energy structure co-evolution signal.

[0023] Furthermore, the calculation logic of the energy structure co-evolution signal includes: using a nonlinear function to calculate the contribution rate of the initial mechanical energy input of the anode carbon block to the generation of the final product structure; using a dynamic correction operator to compensate for environmental factors and eliminate time delay interference; and synthesizing and outputting a signal characterizing the energy efficiency conversion potential of the entire chain.

[0024] Furthermore, the execution of the linkage compensation instruction in step S4 includes: when the energy structure co-evolution signal shows that the structural evolution trend is below a threshold, triggering the performance self-healing logic, including front-end compensation and back-end correction, specifically:

[0025] Front-end compensation optimizes micro-packing density by increasing the entropy of gradation topology information in the next batch of raw material proportions.

[0026] The back-end correction is based on the temporal variation of the interface charge transfer impedance, and simultaneously corrects the heating slope and isothermal duration of the current batch during the calcination stage.

[0027] The linkage compensation command also includes an elegant degradation mode for the production stability of the anode carbon block: when it is determined that the feedback data of the anode carbon block in the roasting process is abnormal, it automatically switches to the single-point compensation mode, stops the correction of the roasting curve, and performs single-dimensional performance self-healing by adjusting the "gradation topology information entropy" in the raw material ratio of the next batch of anode carbon blocks.

[0028] A low-power-consumption anode carbon block is prepared by a low-power-consumption anode carbon block preparation method.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] This invention achieves the alignment of the elastic manifold time axis of the current time-frequency characteristics at the minute-level sampling frequency in the kneading process stage with the thermodynamic dissipation characteristics at the day-level sampling frequency in the roasting stage by constructing a virtual batch shadow memory mechanism. The batch identifier alignment index eliminates the "information gap" and feedback lag between processes. By using a residual mapping network, the first feature vector representing the micro-wetting characteristics and the second feature vector representing the thermodynamic characteristics are coupled into a high-dimensional tensor form of a multidimensional microstructure evolution potential energy matrix. By limiting the upper limit of energy transfer through spatial dimension constraint operators and dynamically correcting the evolution rate through time dimension driving operators, a microstructure evolution simulation model representing the dynamic evolution of the anode carbon block from raw block to finished product is established. This achieves a quantitative description of the conversion efficiency of "mechanical energy input" to "final dense structure generation energy" and a precise quantification of the energy loss of the entire chain. By calculating the cross-process performance coupling correlation degree and combining it with the dynamic correction operator to synthesize the energy structure co-evolution signal, the production system has the ability to see through the "hidden defects" inside the carbon block across processes, which significantly improves the prediction accuracy of the finished product quality distribution trend.

[0031] This invention also utilizes a linkage compensation instruction set generated based on co-evolution signal analysis. When the detected structural evolution trend is below the performance self-healing trigger threshold of 0.65, it dynamically increases the entropy of the gradation topology information in the next batch of raw materials to 1.62 to optimize the micro-packing compactness. Simultaneously, it corrects the heating slope and isothermal duration of the current batch in the calcination stage in real time based on the temporal variation of interface charge transfer impedance. This constructs a cross-process performance self-healing closed-loop preparation system with "automatic defect offsetting" function. Under extreme conditions of abnormal sensor feedback, it gracefully reduces... The graded mode increases the gain weight of the front-end gradation compensation from 0.55 to 0.85, ensuring that the production system can maintain extremely high robustness and performance consistency even in the event of hardware failure. The coefficient of variation of the coking interface thickness of the aggregate and binder inside the final anode carbon block is consistently below 0.08, and the uniformity of the bulk density distribution of the finished product is improved by more than 10%. While significantly improving the refractory performance and electrical conductivity consistency, it can reduce the overall power consumption of anode production by more than 15% compared with the traditional solution, and comprehensively solves the production stability problem caused by raw material fluctuations and process time delays. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0033] Figure 2 This is a schematic diagram of the overall process framework of S3 and S4 in this invention. Detailed Implementation

[0034] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0035] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0036] Please see Figures 1 to 2 This invention provides a technical solution: a method for preparing low-power-consumption anode carbon blocks, the specific steps of which include:

[0037] S1. Extract the first feature vector characterizing the aggregate binder wetting characteristics of the anode carbon block to be prepared during the kneading process, and the second feature vector characterizing the thermodynamic characteristics of the anode carbon block during the calcination process.

[0038] S2. Construct the multidimensional microstructure evolution potential energy matrix of the anode carbon block, use the first eigenvector as the spatial dimension constraint operator and the second eigenvector as the time dimension driving operator, and establish a microstructure evolution simulation model that characterizes the dynamic evolution of the anode carbon block from raw block to finished product.

[0039] S3. Based on the microstructure evolution simulation model, perform spatiotemporal scale alignment operation based on batch identifier to generate energy structure co-evolution signal that quantifies the energy efficiency conversion loss of the entire chain;

[0040] S4. Analyze the energy structure co-evolution signal, calculate the linkage compensation command for the current batch, and adjust the raw material ratio parameters and calcination temperature gradient curve of the next batch accordingly to achieve closed-loop preparation of anode carbon blocks with cross-process performance feedback compensation.

[0041] In step S1, the process of extracting the first feature vector includes:

[0042] Acquire the current timing data of the kneading drive motor during the kneading process; perform a sliding window Fourier transform on the current timing data to calculate the power consumption spectral density statistical moments under different time windows; and calculate the porous medium wave impedance evolution rate, which characterizes the energy transfer efficiency inside the anode carbon block material, based on the power consumption spectral density statistical moments.

[0043] Using the power spectral entropy of the current time series data in the frequency domain, the microscopic wetting degree at the interface between the aggregate and the binder during the mixing process of the anode carbon block is quantified, and the microscopic wetting degree is used as the weight gain of the wave impedance evolution rate of the porous medium to construct the first feature vector.

[0044] In S1, the process of extracting the second feature vector includes acquiring the thermodynamic multi-source sensing data of the anode carbon block in the roasting process, wherein the thermodynamic multi-source sensing data includes the furnace temperature gradient time sequence signal and the flue gas flow channel heat flux density signal.

[0045] The thermodynamic energy efficiency distribution entropy of the thermodynamic multi-source sensing data is calculated to characterize the energy dissipation characteristics of the anode carbon block during the thermodynamic evolution process; the thermal strain energy density variational characteristics of the anode carbon block at different calcination temperature ranges are extracted, and the second feature vector is constructed by combining the thermodynamic energy efficiency distribution entropy.

[0046] In step S2, the spatial dimension constraint operator specifically executes the following logic: using the first eigenvector as the boundary condition mapping parameter of the potential energy matrix to limit the upper limit of energy transfer at different grid points; the time dimension driving operator specifically executes the following logic: using the second eigenvector as the input weight of the time step sequence to dynamically correct the evolution rate of nonlinear phase transition points in the matrix.

[0047] The process of constructing the multidimensional microstructure evolution potential matrix includes: retaining the original information of the first eigenvector and the second eigenvector through the identity mapping branch, extracting the evolution perturbation of the carbon block structure through the nonlinear mapping branch, and performing element-level addition operation on the original information and the evolution perturbation before performing tensor dimension reshaping.

[0048] The calculation logic of the multidimensional microstructure evolution potential energy matrix includes: establishing the energy distribution mapping relationship of the anode carbon block material during the transformation process from a loose particulate state to a continuous medium state; identifying the potential energy evolution poles of the anode carbon block feature vector at different preparation stages in real time, and using them as input features for nonlinear causal mapping in the simulation model.

[0049] Furthermore, the multidimensional microstructure evolution potential matrix is ​​constructed by performing the following steps:

[0050] The first and second feature vectors, after alignment processing, are obtained and concatenated as column vectors to form the original combined feature space. A preset evolution weight matrix (the configuration file is a parameter set pre-calibrated and stored based on historical sample data) is loaded from an external data carrier, including a preset configuration file. Feature extraction is performed using a residual mapping operator, specifically including: taking the original combined feature space as input, generating a nonlinear structural evolution increment through a mapping branch containing a linear transformation layer and a ReLU activation layer, while retaining an identity transmission branch that directly skips the mapping branch; performing an addition operation on the nonlinear structural evolution increment output by the mapping branch and the original feature output by the identity transmission branch to generate an intermediate state feature tensor with residual correction features; mapping the intermediate state feature tensor to a three-dimensional tensor space through a reshaping operator to construct the multidimensional microstructure evolution potential energy matrix characterizing the energy distribution inside the anode carbon block.

[0051] The current sampling frequency parameter of the kneading drive motor is obtained through an industrial data acquisition card. This parameter is denoted as fsa, and in this embodiment, its specific value is set to 1024 Hz. The technical consideration for setting this value is that when carbon aggregate is subjected to shear force crushing and microscopic impregnation with asphalt binder, it will generate instantaneous high-frequency nonlinear fluctuations. According to the Nyquist sampling theorem, 1024 Hz can completely capture the core feature components in the 200 to 400 Hz range, thereby avoiding information distortion caused by signal aliasing.

[0052] In the specific execution stage of feature extraction, frequency domain depth mining based on sliding window processing is performed on the acquired raw current sequence. The specific calculation logic is as follows: a sliding time window containing 512 sampling points is set; for the current data within the time window, windowed autocorrelation operation is performed to convert the time-domain fluctuation signal into an energy correlation function; a Fourier transform is performed on the correlation function to obtain the power spectral density (PSD) characterizing the energy frequency domain distribution; finally, the statistical moments of the power spectral density (PSD) are calculated, denoted as Mpsd. To quantify the degree of micro-wetting at the aggregate-binder interface during mixing, the power spectral entropy parameter is further calculated, denoted as Hps. The calculation logic for this parameter is as follows: The energy values ​​of each frequency point in the power spectrum are taken and their probability proportions to the total energy are calculated. Each proportion value is multiplied by its own natural logarithm. All product results are summed and the opposite is taken to obtain the entropy value characterizing the disorder of the signal. In this embodiment, the original entropy value is normalized by dividing it by the theoretical maximum entropy value to obtain Hps characterizing the disorder of the signal. Accordingly, the output of Hps is limited to the range [0, 1]. When its value is between 0.8 and 0.9, it physically corresponds to a high-quality wetting state in which the binder has fully penetrated into the micropores of the aggregate.

[0053] Based on the aforementioned frequency domain characteristics, the "porous medium wave impedance evolution rate," which characterizes the energy transfer efficiency within the material, is calculated and denoted by the parameter symbol Dk. Its logic originates from the porous medium mechanical model, used to describe how mechanical energy is transferred through semi-fluid carbon material. This parameter serves as the first feature vector of this invention, and its value is determined by obtaining the instantaneous rate of change of HPs and weighting it with a preset impedance mapping coefficient. When the carbon material transforms from a loose particle group into a continuous viscoelastic medium, Dk exhibits a significant monotonically increasing trend. In this embodiment, the final output of Dk is also limited to the range [0, 1], and the closer its value is to 0.9, the higher the density of the carbon block at the end of the kneading process. The innovation of this step lies in transforming the invisible microscopic physical process into a calculable industrial big data indicator, providing a high-confidence pre-judgment criterion for subsequent simulation design.

[0054] In the simulation model construction phase, this embodiment introduces an improved residual mapping network to construct a multidimensional microstructure evolution potential matrix. Pre-defined evolution weight parameters are loaded from an external data carrier (i.e., the configuration file of the production database), and dimension splicing logic is executed. This logic maps a first eigenvector representing the kneading and wetting characteristics to the same high-dimensional tensor space along with a second eigenvector representing the calcination thermodynamic dissipation characteristics (i.e., the temporal variation of interfacial charge transfer impedance).

[0055] Furthermore, by constructing an adaptive tracking operator for dynamic potential energy poles, the potential energy evolution poles of the anode carbon block feature vector at different preparation stages can be identified in real time. Specifically, the calculation process of the adaptive tracking operator includes obtaining the energy gradient value of each grid point in the potential energy matrix; identifying the critical point with the largest gradient change, i.e., the potential energy evolution pole, through a nonlinear causal mapping model; and adaptively correcting the weight distribution of the matrix based on this pole. The technical consideration for setting this logic is that there is a phase transition point with drastic changes in physical properties during the transformation of the carbon block from a loose state to a continuous medium state. By capturing this pole, the simulation model can predict the risk of microcrack development in the finished product in advance.

[0056] Based on the evolution trend of the potential energy matrix, a co-evolution signal of the energy structure is generated, denoted by the parameter symbol Ssy. The synthesis logic of this signal follows a nonlinear mapping function: calculating the cosine similarity between the first and second eigenvectors in the matrix space; extracting the co-evolution gain deviation value; and converting the deviation value into the cross-process performance coupling correlation degree. When Ssy shows that the structural evolution trend is lower than a preset threshold, the control system will trigger the performance self-healing logic. The execution of this linkage compensation command includes: automatically increasing the "gradation topology information entropy" in the next batch of raw materials at the front end to optimize the packing density, and simultaneously extending the isothermal duration of the roasting process at the back end. This cross-process collaborative mechanism achieves the technical effect of "automatic defect offsetting" through big data mining. Compared with traditional solutions, it can reduce the overall power consumption of anode production by more than 15% and significantly improve the consistency and refractory performance of the finished product.

[0057] The process of performing spatiotemporal scale alignment in step S3 includes: associating the first feature vector with the corresponding production batch identifier and temporarily storing it in the virtual batch shadow memory; when the batch identifier enters the roasting process, retrieving the corresponding first feature vector from the memory to achieve alignment of the kneading feature with the sampling frequency of minutes and the roasting feature with the sampling frequency of days on the simulation time axis.

[0058] The process of generating the energy structure co-evolution signal includes: calculating the cosine similarity between the aligned first feature vector and the second feature vector in the simulation model; extracting the co-gain deviation value and converting it into the cross-process performance coupling correlation degree as a quantitative index of the energy structure co-evolution signal.

[0059] Furthermore, the calculation logic of cosine similarity includes mapping the first feature vector and the second feature vector to a unified dimensionless evaluation space using a projection operator based on manifold learning; performing vector length normalization on the mapped features so that each feature component is in the value range of zero to one; calculating the dot product of the two normalized feature vectors and calculating the product of the magnitudes of each feature vector; calculating the ratio of the dot product to the product of the magnitudes to obtain the cosine similarity value; the cooperative gain deviation value is obtained by obtaining a preset target similarity benchmark value, subtracting the cosine similarity value, and taking the absolute value.

[0060] Furthermore, before performing the cosine similarity calculation, the mechanical power consumption time-frequency feature vector and the thermodynamic dissipation feature vector are preprocessed, including projecting the first feature vector and the second feature vector onto a unified energy efficiency evaluation manifold space through a preset mapping matrix; using Z-score normalization to eliminate the dimensional differences between the ampere time-frequency signal and the temperature gradient signal; and calculating the cosine value of the angle between the two normalized vectors in the energy efficiency evaluation manifold space.

[0061] The calculation logic of the energy structure co-evolution signal includes: using a nonlinear function to calculate the contribution rate of the initial mechanical energy input of the anode carbon block to the generation of the final product structure; using a dynamic correction operator to compensate for environmental factors and eliminate time delay interference; and synthesizing and outputting a signal characterizing the energy efficiency conversion potential of the entire chain.

[0062] Furthermore, in the industrial scenario of anode carbon block preparation, the mechanical power consumption characteristics generated during the kneading process are high-frequency dynamic data on a minute-by-minute scale, while the thermodynamic evolution of the roasting process exhibits long-term, large-hysteresis characteristics on a day-by-day scale. This extreme sampling frequency mismatch and time span discontinuity is the main technical bottleneck for achieving full-chain energy efficiency analysis. This embodiment solves this problem by constructing a virtual batch shadow memory mechanism. A unique batch identifier alignment index is generated for each batch of anode carbon blocks to be processed, with its parameter symbol denoted as Dq. The index is determined by: obtaining the start time stamp of the current kneading pot; extracting the production sequence number of the current batching belt scale; and using a preset hash algorithm to map the time stamp and the sequence number to generate a fixed-length hexadecimal character sequence, including "B20231024001". In this way, a unique key constraint is established in the industrial relational database, ensuring the absolute accuracy of subsequent data retrieval.

[0063] During the kneading process, a virtual space with a fixed addressing period, known as a virtual batch shadow memory, is allocated in the cache. The first feature vector (i.e., the mechanical power consumption time-frequency feature vector) is written into this memory using Dq as a logical pointer for temporary "shadowing." When the charcoal block corresponding to this batch identifier is formed, grouped, and enters the roasting process several days later, the sensor trigger at the roasting furnace inlet reads its corresponding Dq. At this time, an elastic manifold time axis alignment operation is performed: the kneading feature vector from several days ago is retrieved from the shadow memory based on Dq; the deviation coefficient of the current production rhythm is calculated. If this coefficient exceeds a preset threshold (set to 0.1 in this embodiment, representing a production delay fluctuation exceeding 10%), the kneading feature is upsampled using a spline interpolation algorithm. The technical consideration for setting this logic is that, through nonlinear manifold projection, feature misalignment caused by environmental temperature differences or logistics scheduling can be compensated, allowing minute-level kneading data to be accurately aligned to specific coordinate points on the day-level roasting simulation time axis.

[0064] After completing the cross-process feature alignment, the process proceeds to the synthesis stage of the energy structure co-evolution signal. This process is achieved by calculating the similarity between the aligned mechanical power consumption feature vector and the thermodynamic dissipation feature vector in the multidimensional microstructure evolution potential energy matrix space. The specific calculation steps are as follows: normalize the two vectors to eliminate dimensional influence; calculate the cosine value of the angle between these two high-dimensional vectors, i.e., the cosine similarity; extract the co-gain deviation value based on this similarity, denoted by the parameter symbol Xpz; the physical meaning of Xpz lies in quantifying the conversion loss between the initial mechanical power consumption input and the final microstructure formation energy, and its value range is limited to the interval between 0 and 1. In this embodiment, if Xpz is 0.05, it indicates extremely high energy conversion efficiency, meaning that the initial kneading mechanical power consumption can be effectively converted into a denser structure within the carbon block.

[0065] Finally, an environmental factor compensation is applied to the aforementioned synergistic indicators using a dynamic correction operator to synthesize the final energy structure synergistic evolution signal, denoted as Nj. The coefficient determination logic of this dynamic correction operator is as follows: real-time analysis of the residual between the measured resistivity and simulated value of the finished anode over the past three months, and calculation of the perturbation weights of atmospheric humidity and ambient temperature on the energy efficiency contribution rate using the gradient boosting method. The corrected indicators are then integrated into a cross-process performance coupling correlation degree, denoted as Og. The final output Nj signal serves as a quantitative indicator characterizing the energy efficiency conversion potential of the entire carbon block chain. The closer its value is to 0.9, the more optimal the electrical conduction structure evolution trend of this batch of carbon blocks is under the current thermodynamic environment. This entire process achieves cross-process insight into the "hidden defects" within the carbon blocks, providing high-confidence data support for subsequent performance self-healing closed-loop control.

[0066] To verify the effectiveness of the spatiotemporal scale alignment logic and co-evolution signal calculation proposed in this invention, this embodiment constructs a simulation platform for charcoal block preparation simulating a 240-hour continuous production cycle. The specific parameters of the mathematical model are set as follows: the kneading current sampling frequency is set to 1024 Hz, the calcination temperature gradient monitoring period is set to 12 minutes, the shadow memory addressing period is set to 10 milliseconds, and the manifold projection space dimension is set to 128 dimensions.

[0067] Table 1: Validation Table of Simulation Experiment Data on Cross-Process Collaborative Evolution of Anode Carbon Blocks

[0068] Experiment 1 B2405-001 0.02 Normal hit 0.92 0.03 0.94 1.63 Experiment 2 B2405-001 0.03 Normal hit 0.91 0.04 0.92 1.62 Experiment 3 B2405-002 0.14 Interpolation activation 0.88 0.07 0.89 1.59 Experiment 4 B2405-002 0.15 Interpolation activation 0.87 0.08 0.88 1.58 Control group 1 C2405-X01 0.22 Logical disabling 0.58 0.25 0.52 1.48 Control group 2 C2405-X02 0.25 Logical disabling 0.55 0.28 0.49 1.46

[0069] To quantify and mitigate time-lag interference during the production process, this embodiment of the invention introduces a production rhythm deviation coefficient (Cdev) during feature alignment. The logic for determining the production rhythm deviation coefficient (Cdev) is as follows: the data processing module calculates the actual production cycle by obtaining the difference between the trigger time stamp of the current roasting process and the kneading completion time stamp temporarily stored in the virtual batch shadow memory; subsequently, it calculates the absolute value of the difference between the actual production cycle and the preset standard process cycle; finally, it divides this difference by the preset standard process cycle to obtain the production rhythm deviation coefficient reflecting the time dimension. Physically, this parameter quantitatively characterizes the degree of time lag between processes caused by logistics scheduling, environmental temperature differences, or machine downtime due to malfunctions, serving as the core criterion for triggering the shadow memory to perform elastic manifold time axis alignment operations.

[0070] Analysis of simulation data from a 240-hour continuous production cycle shows that when the production rhythm deviation coefficient (Cdev) is in the high range of 0.14 to 0.25 (i.e., severe fluctuations in production rhythm), the experimental group executing virtual batch shadow memory retrieval and elastic manifold alignment logic still maintains a stable inter-process performance coupling correlation (ICI) above 0.88. In contrast, the control group with the alignment logic disabled experiences a rapid drop in its inter-process performance coupling correlation (ICI) below 0.52. Data comparison calculations show that, when facing the same intensity of production disturbance, the experimental group of this invention achieves an average improvement of approximately 73.08% in inter-process performance coupling correlation (ICI) compared to the control group (calculated based on: This definitively proves that the solution has extremely high accuracy in capturing cross-process evolution trends.

[0071] In terms of verifying the physical properties of the finished product, the measured bulk density of the anode carbon block prepared in Experimental Group 3 reached 1.59 g / cm³. 3 The control group 1, which did not employ the alignment logic of this invention, had a concentration of only 1.48 g / cm³. 3The data results show that, by performing spatiotemporal scale alignment operations based on batch identifiers, this invention achieves a significant improvement of approximately 7.43% in the finished product density index even under conditions of large time-lag fluctuations (calculation basis: The experimental data above not only verify the feasibility of the present invention in cross-process feature retrieval and collaborative mapping technology, but also quantitatively reveal its outstanding substantive features and significant technological progress compared with existing technologies in filtering industrial production environment disturbances and maintaining a high degree of consistency in finished product quality.

[0072] The execution of the linkage compensation command in step S4 includes: when the energy structure co-evolution signal shows that the structural evolution trend is below a threshold, triggering the performance self-healing logic, including front-end compensation and back-end correction, specifically:

[0073] Front-end compensation optimizes micro-packing density by increasing the entropy of gradation topology information in the next batch of raw material proportions.

[0074] The back-end correction is based on the temporal variation of the interface charge transfer impedance, and simultaneously corrects the heating slope and isothermal duration of the current batch during the calcination stage.

[0075] The linkage compensation command also includes an elegant degradation mode for the production stability of the anode carbon block: when it is determined that the feedback data of the anode carbon block in the roasting process is abnormal, it automatically switches to the single-point compensation mode, stops the correction of the roasting curve, and performs single-dimensional performance self-healing by adjusting the "gradation topology information entropy" in the raw material ratio of the next batch of anode carbon blocks.

[0076] Furthermore, by executing the structural self-healing and linkage control logic in step S4, the energy structure co-evolution signal output in step S3 is analyzed and transformed into a linkage compensation instruction set that directly drives the production terminal. Through preventive compensation of the front-end raw material ratio and real-time correction of the back-end roasting curve, the electrical conduction consistency of the finished anode in a complex production environment is ensured.

[0077] Specifically, a deep analysis is performed on the energy structure co-evolution signal, and a key judgment parameter, namely the performance self-healing trigger threshold, is introduced. This parameter is denoted as Xn. Its function is to determine whether cross-process intervention logic needs to be initiated. In this embodiment, the value of Xn is set to 0.65. The technical consideration for setting this value is that, through historical production big data mining, it was found that when the co-evolution signal is below this value, the microscopic defects of the finished carbon blocks will cause the resistivity to exceed the control boundary of industrial grade one products. The calculation logic is as follows: obtain the current value of the co-evolution signal generated in step S3; perform a subtraction operation between this signal value and the threshold Xn to obtain a deviation value; determine the sign of this deviation value. If the deviation value is negative, it is determined that the current batch of carbon blocks has a performance decline trend, and the linkage compensation process is immediately triggered.

[0078] In the front-end compensation stage, the packing density defect of the current batch is offset by adjusting the "gradation topological information entropy" in the raw material ratio of the next batch of carbon blocks. The gradation topological information entropy (parameter symbol Jp) is a core parameter characterizing the spatial packing potential of aggregate particle size distribution. It is obtained based on the Shannon entropy model in information theory, and the specific calculation process is broken down as follows: The mass proportion of each aggregate size is collected in real time by sensors in the batching system; each proportion value is multiplied by its own natural logarithm to obtain multiple intermediate products; the intermediate products of all particle sizes are summed by addition; and the sum is inversely represented to obtain the instantaneous gradation topological information entropy. In this embodiment, when a performance deviation is detected, the target value of Jp for the next batch is dynamically increased from the baseline value of 1.45 to 1.62. The technical rationale for this is that by increasing the filling weight of fine particles in the topological space, a denser aggregate bond can be formed at the microscopic level, thereby offsetting the performance gap of the "historical batch" by improving the structural strength of the "future batch".

[0079] In the back-end correction stage, the calcination temperature rise curve correction based on the "temporal variation degree of volatile matter pyrolysis reaction rate" is initiated simultaneously; the impedance temporal variation degree (parameter symbol Jdh) is constructed based on the coupled analysis logic of in-furnace gas component monitoring and local temperature field thermal effects, and is used to quantify the intensity of volatile matter release and pore structure reorganization during the coking process of binder asphalt; the specific implementation steps are as follows:

[0080] In the carbonization-sensitive range of the roasting process (set to 300°C to 600°C in this embodiment), a high-temperature thermocouple array and a flue gas analysis probe are used to synchronously collect the furnace pit flue temperature and the temperature rise deviation data generated by volatile combustion at a preset minute-level frequency; the temperature rise rate deviation value between adjacent sampling points is calculated; the temperature rise rate deviation value is normalized and used as an instantaneous rate characterizing the intensity of the internal reaction; a weighted integral processing is performed on this rate to extract the variation index reflecting the evolution trend;

[0081] In this embodiment, the output value of Jdh is limited to the range of 0 to 1. If the impedance timing variation exceeds the preset safety limit (e.g., 0.8), it indicates that the rapid precipitation of volatiles inside the carbon block leads to an excessively violent pyrolysis reaction (which is prone to causing cracks). The following instructions will be automatically issued: obtain the original heating slope; calculate the variation deviation, i.e., the current variation minus the safety threshold; multiply the original heating slope by the difference between "1 minus the variation deviation" to obtain the corrected gentle heating slope; and simultaneously extend the current isothermal duration. The extension time is determined by the product of the variation deviation and the reference duration (set to 120 minutes in this embodiment).

[0082] By introducing a dynamic error inverse compensation gain operator, the compensation characteristics of the front end and the back end are integrated to further improve the synergistic effect of the linkage control, specifically including:

[0083] The algorithm acquires the incremental value of the topological entropy of the gradation; obtains the compensation amount for the calcination isothermal duration; calculates the synergistic gain coefficient of the two using a dimensionality reduction fusion algorithm based on principal component analysis; and dynamically adjusts the execution intensity of the linkage command based on the synergistic gain coefficient. This fusion mechanism can avoid overload of single-process adjustment, ensuring performance self-healing while maximizing the reduction of calcination energy consumption.

[0084] Furthermore, this embodiment also features a graceful degradation mode for production stability. When the self-test detects a signal interruption or data overflow in the impedance sensor of the calcination section (i.e., the signal is at the upper limit of the range for three consecutive cycles), the logic judgment module automatically cuts off the back-end correction loop. At this time, the following switching logic is executed: the current calcination temperature rise curve is locked, and dynamic adjustment is stopped; the gain weight of the front-end gradation compensation is increased from the initial 0.55 to 0.85; and single-dimensional performance self-healing is performed only by adjusting the gradation topology information entropy of the next batch. This design ensures that even under extreme conditions of sensor hardware failure, the production system can still rely on big data prediction logic to maintain a minimum level of performance mitigation, avoiding quality accidents caused by erroneous feedback, demonstrating the high robustness of the control system. The final output linkage instruction set consists of the frequency control signal of the electronic belt scale and the air-fuel ratio adjustment instruction of the calcination furnace burner, realizing a closed-loop preparation of anode carbon blocks with cross-process performance self-healing.

[0085] A low-power-consumption anode carbon block is prepared by a low-power-consumption anode carbon block preparation method.

[0086] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max-normalization and Z-score standardization.

[0087] The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random-access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0088] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for preparing low-power-consumption anode carbon blocks, characterized in that, The specific steps include: S1. Extract the first feature vector characterizing the aggregate binder wetting characteristics of the anode carbon block to be prepared during the kneading process, and the second feature vector characterizing the thermodynamic characteristics of the anode carbon block during the calcination process. S2. Construct the multidimensional microstructure evolution potential energy matrix of the anode carbon block, use the first eigenvector as the spatial dimension constraint operator and the second eigenvector as the time dimension driving operator, and establish a microstructure evolution simulation model that characterizes the dynamic evolution of the anode carbon block from raw block to finished product. S3. Based on the microstructure evolution simulation model, perform spatiotemporal scale alignment operation based on batch identifier to generate energy structure co-evolution signal that quantifies the energy efficiency conversion loss of the entire chain; S4. Analyze the energy structure co-evolution signal, calculate the linkage compensation command for the current batch, and adjust the raw material ratio parameters and calcination temperature gradient curve of the next batch accordingly to achieve closed-loop preparation of anode carbon blocks with cross-process performance feedback compensation.

2. The method for preparing a low-power-consumption anode carbon block according to claim 1, characterized in that: In step S1, the process of extracting the first feature vector includes: Acquire the current timing data of the kneading drive motor during the kneading process; perform a sliding window Fourier transform on the current timing data to calculate the power consumption spectral density statistical moments under different time windows; and calculate the porous medium wave impedance evolution rate, which characterizes the energy transfer efficiency inside the anode carbon block material, based on the power consumption spectral density statistical moments. Using the power spectral entropy of the current time series data in the frequency domain, the microscopic wetting degree at the interface between the aggregate and the binder during the mixing process of the anode carbon block is quantified, and the microscopic wetting degree is used as the weight gain of the wave impedance evolution rate of the porous medium to construct the first feature vector.

3. The method for preparing a low-power-consumption anode carbon block according to claim 2, characterized in that: In S1, the process of extracting the second feature vector includes acquiring the thermodynamic multi-source sensing data of the anode carbon block in the roasting process. The thermodynamic multi-source sensing data includes the furnace temperature gradient time sequence signal and the flue gas flow channel heat flux density signal. The thermodynamic energy efficiency distribution entropy of the thermodynamic multi-source sensing data is calculated to characterize the energy dissipation characteristics of the anode carbon block during the thermodynamic evolution process. The thermal strain energy density variational characteristics of the anode carbon block under different calcination temperature ranges are extracted, and the second feature vector is constructed by combining it with the thermodynamic energy efficiency distribution entropy.

4. The method for preparing a low-power-consumption anode carbon block according to claim 3, characterized in that: In step S2, the spatial dimension constraint operator specifically executes the following logic: using the first eigenvector as the boundary condition mapping parameter of the potential energy matrix to limit the upper limit of energy transfer at different grid points; the time dimension driving operator specifically executes the following logic: using the second eigenvector as the input weight of the time step sequence to dynamically correct the evolution rate of nonlinear phase transition points in the matrix. The process of constructing the multidimensional microstructure evolution potential matrix includes: retaining the original information of the first eigenvector and the second eigenvector through the identity mapping branch, extracting the evolution perturbation of the carbon block structure through the nonlinear mapping branch, and performing element-level addition operation on the original information and the evolution perturbation before performing tensor dimension reshaping.

5. The method for preparing a low-power-consumption anode carbon block according to claim 4, characterized in that: The calculation logic of the multidimensional microstructure evolution potential energy matrix includes: establishing the energy distribution mapping relationship of the anode carbon block material during the transformation process from a loose particulate state to a continuous medium state; identifying the potential energy evolution poles of the anode carbon block feature vector at different preparation stages in real time, and using them as input features for nonlinear causal mapping in the simulation model.

6. The method for preparing a low-power-consumption anode carbon block according to claim 5, characterized in that: The process of performing spatiotemporal scale alignment in step S3 includes: associating the first feature vector with the corresponding production batch identifier and temporarily storing it in the virtual batch shadow memory; when the batch identifier enters the roasting process, retrieving the corresponding first feature vector from the memory to achieve alignment of the kneading feature with the sampling frequency of minutes and the roasting feature with the sampling frequency of days on the simulation time axis.

7. The method for preparing a low-power-consumption anode carbon block according to claim 6, characterized in that: The process of generating the energy structure co-evolution signal includes: calculating the cosine similarity between the aligned first feature vector and the second feature vector in the simulation model; extracting the co-gain deviation value and converting it into the cross-process performance coupling correlation degree as a quantitative index of the energy structure co-evolution signal.

8. The method for preparing a low-power-consumption anode carbon block according to claim 7, characterized in that: The calculation logic of the energy structure co-evolution signal includes: using a nonlinear function to calculate the contribution rate of the initial mechanical energy input of the anode carbon block to the generation of the final product structure; using a dynamic correction operator to compensate for environmental factors and eliminate time delay interference; and synthesizing and outputting a signal characterizing the energy efficiency conversion potential of the entire chain.

9. The method for preparing a low-power-consumption anode carbon block according to claim 8, characterized in that: The execution of the linkage compensation command in step S4 includes: when the energy structure co-evolution signal shows that the structural evolution trend is below a threshold, triggering the performance self-healing logic, including front-end compensation and back-end correction, specifically: Front-end compensation optimizes micro-packing density by increasing the entropy of gradation topology information in the next batch of raw material proportions. The back-end correction is based on the temporal variation of the interface charge transfer impedance, and simultaneously corrects the heating slope and isothermal duration of the current batch during the calcination stage. The linkage compensation command also includes an elegant degradation mode for the production stability of the anode carbon block: when it is determined that the feedback data of the anode carbon block in the roasting process is abnormal, it automatically switches to the single-point compensation mode, stops the correction of the roasting curve, and performs single-dimensional performance self-healing by adjusting the "gradation topology information entropy" in the raw material ratio of the next batch of anode carbon blocks.

10. A method for preparing a low-power-consumption anode carbon block, characterized in that: The finished anode carbon block is prepared by a low-power consumption anode carbon block preparation method according to any one of claims 1-9.

Images