Aluminum electrolysis cell ledge dynamic prediction method and device, storage medium and product

Abstract

The invention discloses an aluminum electrolysis cell ledge dynamic prediction method and device, a storage medium and a product, and the method comprises the steps: constructing an aluminum electrolysis cell multi-physical field simulation model according to structure parameters and process parameters, obtaining ledge thicknesses and cell shell temperatures under different process parameters, and constructing a sample data set; training the ledge size prediction model by using the sample data set to obtain a target prediction model; predicting the real-time process parameters of the aluminum electrolysis cell to be predicted by using the target prediction model to obtain a ledge thickness prediction value and a cell shell temperature prediction value; judging whether the target prediction model meets the prediction precision requirement or not according to the ledge thickness measured value and the ledge thickness predicted value, and adjusting the aluminum electrolysis cell multi-physical field simulation model and retraining the target prediction model according to the judgment result until the prediction precision requirement is met. According to the method, the prediction precision and robustness are improved, and the adaptability to complex and changeable working conditions is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of aluminum electrolysis cells, and in particular relates to a method, device, storage medium and product for dynamically predicting the ridge of an aluminum electrolysis cell. Background Art

[0002] Aluminum electrolytic cells are key electrolytic equipment in the aluminum metallurgical industry, used to extract aluminum from alumina through an electrometallurgical process. The main components of aluminum electrolytic cells include cell shells, cell chambers, anodes, cathodes, linings, and electrolytes. The performance and service life of the electrolytic cell are of paramount importance in the electrolytic process and are affected by multiple factors, among which the shape, thickness, and thermal stability of the cell sides directly determine the operating efficiency and safety of the electrolytic cell. The cell sides mainly refer to the two side areas of the electrolytic cell, which are the areas most exposed to heat, stress, and corrosion during the operation of the electrolytic cell.

[0003] With the development of aluminum electrolysis technology, the size of aluminum electrolytic cells has gradually increased (the current maximum has reached 600kA), and the operating conditions have become more complex and changeable, which has put forward higher requirements on the thermal behavior, structural changes and life of the cell side. Therefore, how to accurately predict and monitor the dynamic changes of the cell side, timely detect the abnormal state of the cell side and perform fault diagnosis, and then provide certain feedback for the control of the electrolytic cell, has become an important technical problem in the aluminum electrolysis production process.

[0004] The existing technology mainly focuses on estimating the thickness of the tank wall at the point of temperature measurement on the side tank shell, but is not linked to the big data of the electrolytic cell. That is, it is not suitable for prediction under variable working conditions and can only provide simple damage warnings. It is unable to perform real-time high-precision calculation of the thickness of the tank wall on the side of the electrolytic cell, and has poor practical value.

[0005] A Chinese patent document with authorization announcement number CN106709149B discloses a real-time prediction method and system for the three-dimensional furnace shape of an aluminum electrolytic cell based on a neural network. The method mainly calculates the three-dimensional trough side based on a BP neural network model. The method emphasizes the fitting of the neural network model to some process parameters and finite element simulation models, and can only cover a few application scenarios. Summary of the invention

[0006] The purpose of the present invention is to provide a method, device, storage medium and product for dynamically predicting the thickness of the aluminum electrolytic cell, so as to solve at least one of the problems that the traditional method cannot dynamically predict the thickness of the cell wall in real time, has poor prediction accuracy and cannot be applied to complex and changeable working conditions.

[0007] The present invention solves the above technical problems through the following technical solutions: a method for dynamically predicting the ridge of an aluminum electrolytic cell, comprising:

[0008] Acquire structural parameters and process parameters of the aluminum electrolysis cell, and construct a multi-physical field simulation model of the aluminum electrolysis cell according to the structural parameters and process parameters;

[0009] Based on the multi-physics field simulation model of the aluminum electrolytic cell, the cell wall thickness and cell shell temperature under different process parameters are obtained;

[0010] Construct a sample data set based on the groove side thickness and groove shell temperature under different process parameters;

[0011] Constructing a groove side dimension prediction model, and using the sample data set to train the groove side dimension prediction model to obtain a target prediction model;

[0012] Acquire the real-time process parameters of the aluminum electrolytic cell to be predicted, and use the target prediction model to predict the real-time process parameters to obtain a predicted value of the cell side thickness and a predicted value of the cell shell temperature;

[0013] Obtaining a measured value of the thickness of a certain anode of an aluminum electrolytic cell to be predicted;

[0014] According to the groove side thickness measurement value and the groove side thickness prediction value, it is judged whether the target prediction model meets the prediction accuracy requirement; if so, the groove side thickness prediction value and the groove shell temperature prediction value under the real-time process parameters are output; if not, the physical performance parameters of the aluminum electrolytic cell multi-physical field simulation model are adjusted, and based on the adjusted aluminum electrolytic cell multi-physical field simulation model, the steps of constructing the sample data set, retraining the target prediction model, predicting the real-time process parameters and judging the prediction accuracy of the target prediction model are repeated.

[0015] Furthermore, the structural parameters include geometric dimensions and physical performance parameters, the physical performance parameters include the electrical conductivity of the materials in each part and the thermal conductivity under different temperatures and different service life stage working conditions; the process parameters include temperature field, current distribution, aluminum level and electrolyte level.

[0016] Furthermore, the nanoCT method is used to obtain the electrical conductivity and thermal conductivity of the material. The specific implementation process is as follows:

[0017] Avizo software was used to process the acquired CT material data to obtain a three-dimensional sampling model;

[0018] Repairing the three-dimensional sampling model in the model processing software to obtain a repaired model;

[0019] Importing the repaired model into a mesh drawing software, and meshing the repaired model to obtain a material mesoscopic model;

[0020] The material mesoscopic model is imported into finite element simulation software, and the temperature boundary conditions and thermal conductivity boundary conditions are loaded into the material mesoscopic model to calculate the comprehensive electrical conductivity and thermal conductivity of the material.

[0021] Furthermore, based on the multi-physics field simulation model of the aluminum electrolysis cell, the cell wall thickness and cell shell temperature under different process parameters are obtained, specifically including:

[0022] The process parameters, physical boundary conditions and parameter constraints are loaded on the multi-physical field simulation model of the aluminum electrolysis cell, and an iterative solution is performed to obtain the cell wall thickness and cell shell temperature under the process parameters.

[0023] Furthermore, the physical boundary conditions and parameter constraints are optimized by using a multi-scale simulation method.

[0024] Furthermore, a sample data set is constructed based on the groove side thickness and groove shell temperature under different process parameters, including:

[0025] Performing denoising and missing fill processing on the process parameters;

[0026] Standardize and normalize the processed process parameters;

[0027] Integration and format conversion of normalized process parameters;

[0028] Statistical analysis and machine learning techniques are used to extract key features from the process parameters after format conversion, and deep learning algorithms are used to evaluate the correlation between each key feature and the groove side thickness and groove shell temperature.

[0029] The relevant features are determined according to the evaluation results, and a sample data set is constructed based on the determined relevant features and the corresponding groove side thickness and groove shell temperature.

[0030] Furthermore, during the training process of the groove side dimension prediction model or the target prediction model, an automatic parameter tuning algorithm based on Bayesian optimization is used to optimize model parameters.

[0031] Based on the same concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program / instructions stored in the memory, wherein the processor executes the computer program / instructions to implement the aluminum electrolysis cell slab dynamic prediction method as described above.

[0032] Based on the same concept, the present invention also provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the aluminum electrolysis cell slab dynamic prediction method as described above.

[0033] Based on the same concept, the present invention also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the aluminum electrolysis cell slab dynamic prediction method as described above.

[0034] Beneficial Effects

[0035] Compared with the prior art, the advantages of the present invention are:

[0036] The present invention predicts the trough side thickness and trough shell temperature in real time based on real-time process parameters and deep learning technology, and judges whether the target prediction model is suitable for the current working conditions based on the trough side thickness measurement value. By adjusting the multi-physical field simulation model of the aluminum electrolytic cell and retraining the target prediction model, the target prediction model can adapt to the prediction of trough side thickness and trough shell temperature under complex and changeable working conditions, thereby improving the prediction accuracy and robustness, and improving the adaptability to complex and changeable working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only one embodiment of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 1 is a flow chart of a method for dynamically predicting the ridge of an aluminum electrolytic cell according to an embodiment of the present invention;

[0039] Figure 2 : is a CFD simulation result diagram of the inner shape of the groove side under different pole pitches in an embodiment of the present invention;

[0040] Figure 3 It is a visualization diagram of the groove side dynamic prediction result in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following is a clear and complete description of the technical solutions in the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0043] Example 1

[0044] Figure 1The flowchart of the method for dynamic prediction of the aluminum electrolytic cell sidewall provided by the present invention is shown as follows: Figure 1 As shown, the aluminum electrolysis cell ridge dynamic prediction method comprises the following steps:

[0045] Step 1: Obtain the structural parameters and process parameters of the aluminum electrolysis cell.

[0046] In this embodiment, the structural parameters include geometric dimensions and physical performance parameters. The physical performance parameters include the conductivity of the materials of each part and the thermal conductivity under different temperatures and different service life stage working conditions; the process parameters include temperature field, current distribution, aluminum level and electrolyte level. The temperature field includes the side tank shell temperature, the bottom tank shell temperature, and the cathode steel bar temperature. The current distribution includes the anode current. The process parameters of different aluminum electrolytic cells can be acquired through corresponding sensor collection (for example, using distributed thermocouple sensors to collect temperature data, once every 5 seconds), and / or calling historical working condition data stored in SQL database or HDF5 to provide multi-working condition input for the tank side size prediction model. The process parameters can also be acquired by manual collection, such as aluminum level and electrolyte level. When acquired by manual collection, the manually collected data is automatically corrected based on the machine learning model to eliminate subjective errors. The data collected by the sensor can be combined with the historical working condition data and stored in HDF5 format, and gzip compression can be added to reduce the amount of data.

[0047] Taking a 420kA aluminum electrolytic cell as an example, its structural parameters are shown in Table 1.

[0048] Table 1 Some structural parameters of a 420kA aluminum electrolytic cell

[0049] parameter Numeric Remark Series current / kA 420 Adjustable range: 400kA~430kA Anode carbon block size / mm 1700×660×540 Size of each carbon block Cathode carbon block size / mm 3680×665×485 Size of each carbon block Anode carbon block quantity 48 Even distribution Number of cathode carbon blocks 24 Even distribution Cathode steel rod size / mm 2200×100×200 Single steel bar size Number of steel bars 96 The layout of each steel rod Aluminum level / mm 220 Dynamic changes during operation, real-time calculation is required Electrolyte level / mm 180 Dynamic changes during operation require real-time calculation Pole distance / mm 45 Dynamic range: 20~50mm

[0050] In another specific embodiment of the present invention, the operating parameters of the aluminum electrolytic cell can also be obtained (collected once every 10 minutes), including the set voltage, real-time voltage, bus displacement, furnace bottom pressure drop, number of feedings, fluoride salt addition, temperature and needle vibration, etc., and the basic working state of the aluminum electrolytic cell can be obtained through this information. Some operating parameters of the aluminum electrolytic cell are shown in Table 2.

[0051] Table 2 Some operating parameters of aluminum electrolysis cell

[0052] Molecular ratio Fe content Si content Bottom pressure drop Alumina concentration … swing Needle Vibration Mean 2.46 0.10 0.04 315.57 2.96 … 1.23 4.68 Standard Deviation 0.09 0.01 0.00 37.76 0.77 … 0.51 2.13 Minimum 2.17 0.07 0.02 216.00 0.49 … 1.00 1.00 25％ 2.41 0.10 0.03 289.00 2.40 … 1.00 3.00 50％ 2.45 0.10 0.04 310.00 2.85 … 1.00 4.00 75％ 2.50 0.11 0.04 337.00 3.44 … 1.00 6.00 Maximum 3.12 0.29 0.12 495.00 6.22 … 6.00 23.00 Coefficient of variation 0.04 0.18 0.21 0.12 0.26 … 0.41 0.45

[0053] In order to facilitate calling and consulting, all parameters obtained in step 1 are stored in a SQL database.

[0054] Step 2: Construct a multi-physics field simulation model of the aluminum electrolysis cell based on the structural parameters and process parameters obtained in step 1.

[0055] In a specific embodiment of the present invention, the specific implementation process of constructing a multi-physical field simulation model of an aluminum electrolysis cell includes:

[0056] A three-dimensional model of the aluminum reduction cell is constructed according to the structural parameters of the aluminum reduction cell, and then the three-dimensional model is imported into the Openfoam platform for meshing, and material properties are assigned to the meshed model according to the physical performance parameters of different parts.

[0057] In order to improve the accuracy of the physical performance parameters of the materials in each part, the present invention adopts the nanoCT method to obtain the conductivity of the material. The specific implementation process is as follows:

[0058] Step 2.1: Using Avizo software to process the acquired CT material data (i.e., material data acquired through computer tomography) to obtain a three-dimensional sampling model;

[0059] Step 2.2: Repair the 3D sampling model in the model processing software (such as Geomagic Studio), process unreasonable holes, abnormal surfaces and noise points, and obtain the repaired model;

[0060] Step 2.3: Import the repaired model into mesh drawing software (such as fluent meshing), and mesh the repaired model to obtain the material mesoscopic model;

[0061] Step 2.4: Import the material mesoscopic model into the finite element simulation software, and load the temperature boundary conditions and thermal conductivity boundary conditions into the material mesoscopic model to simulate the heat transfer and loss mechanism in different parts and calculate the comprehensive electrical conductivity and thermal conductivity of the material.

[0062] Taking the comprehensive electrical conductivity of the material as the electrical conductivity parameter of the material, the dynamic evolution process of the material property parameters that affect the prediction of the groove side thickness and the groove shell temperature is described by the nanoCT method. The mesoscopic material properties are described in more detail, and the calculation accuracy of the groove side thickness and the groove shell temperature under non-standard working conditions is improved.

[0063] The nanoCT method is an existing technology. For details, please refer to the literature: Wang, Gang, et al. "Quantitative analysis of microscopic structure and gas seepage characteristics of low-rank coal based on CT three-dimensional reconstruction of CT images and fractal theory." Fuel 256(2019):115900.

[0064] Step 3: Based on the multi-physics field simulation model of the aluminum electrolytic cell constructed in step 2, the cell wall thickness and cell shell temperature under different process parameters are obtained.

[0065] In a specific embodiment of the present invention, based on the multi-physics field simulation model of the aluminum electrolysis cell, the cell wall thickness and cell shell temperature under different process parameters are obtained, specifically including:

[0066] The process parameters, physical boundary conditions and parameter constraints are loaded on the multi-physics field simulation model of the aluminum electrolytic cell, and iterative solutions are performed to obtain the thickness of the trough wall and the temperature of the trough shell under the process parameters. The multi-physics field includes the electric field, thermal field and flow field. The electric field and the thermal field interact dynamically, and then combined with the heat conduction equation and the electromagnetic field equation, the heat loss of the side wall of the aluminum electrolytic cell and the trough erosion process are simulated, and the trough wall thickness and the trough shell temperature are calculated.

[0067] Generally, physical boundary conditions and parameter constraints are set based on experience. The present invention uses a multi-scale simulation method to optimize the physical boundary conditions and parameter constraints set based on experience, and calculates the interface behavior of the electrolyte and the inner village material by analyzing the physical and chemical properties of the microstructure. The aluminum electrolysis cell multi-physics field simulation model provides better physical boundary conditions and parameter constraints for the aluminum electrolysis cell multi-physics field simulation model. The multi-scale simulation method is a prior art.

[0068] The multi-physics field simulation model of the aluminum electrolysis cell of the present invention is obtained through microscopic interface behavior, mesoscopic material properties and macroscopic multi-scale modeling methods. It can comprehensively describe the dynamic evolution process of several important parameters affecting the prediction of the cell side, and provide analysis of issues such as cell side solidification and melting, multi-layer heat transfer and full life cycle parameter coverage of electrolyte components.

[0069] Figure 2 The CFD simulation results of the inner shape of the groove side under different pole distances are shown. The numbers on the color ribbon represent the liquid volume fraction, and the solid and liquid phases of the electrolyte are distinguished by the liquid volume fraction.

[0070] Step 4: Construct a sample data set based on the groove side thickness and groove shell temperature under different process parameters obtained in step 3.

[0071] In a specific embodiment of the present invention, a sample data set is constructed according to the groove side thickness and groove shell temperature under different process parameters, including:

[0072] Step 4.1: De-noise and fill in missing information on process parameters to ensure data integrity and accuracy.

[0073] For missing or abnormal values, an interpolation method based on variational autoencoder (VAE) is used to repair them. The interpolation formula is:

[0074]

[0075] Among them, x i Indicates the inserted data, x k 、x j Represents two known data.

[0076] Step 4.2: Standardize and normalize the processed process parameters to eliminate the dimension effect and facilitate subsequent analysis.

[0077] Step 4.3: Integration and format conversion of normalized process parameters; use ETL (extraction, transformation, and loading) process for integration to provide a global perspective for subsequent analysis; convert data from different sources into a unified format (including but not limited to .npy, HDF5), and use built-in compression (such as gzip or zlib) to reduce storage space.

[0078] Depending on the type and source of the data, you can choose to store the data as multiple files, or store the data in different data sets of the same file. For unstructured data, use image formats .png, .jpg or text formats .csv for storage, and combine them with compression technology to improve storage efficiency, reduce space usage, and optimize data access and processing performance.

[0079] Step 4.4: Utilize statistical analysis and machine learning techniques to extract key features from the process parameters after format conversion, and use a deep learning algorithm to evaluate the correlation between each key feature and the groove side thickness and groove shell temperature (e.g., calculate the correlation coefficient).

[0080] Machine learning techniques include but are not limited to principal component analysis, feature importance assessment, etc., and deep learning algorithms include but are not limited to convolutional neural networks and graph neural networks.

[0081] Step 4.5: Determine the relevant features based on the evaluation results, and construct a sample data set based on the determined relevant features and the corresponding groove side thickness and groove shell temperature. That is, select features that have a strong correlation with the groove side thickness and groove shell temperature calculation according to the correlation size to improve the effectiveness and reliability of the prediction model. Each sample in the sample data set includes an input sample and an output sample. The input sample is the relevant feature corresponding to the process parameter, and the output sample is the groove side thickness and groove shell temperature.

[0082] In order to cover more complex working conditions, the distributed current intensity, cover material thickness, electrolyte crystallization temperature, electrode distance and temperature gradient of the aluminum electrolysis cell are also introduced into the input sample.

[0083] Step 5: Construct a groove side dimension prediction model.

[0084] In this embodiment, the groove side size prediction model uses a convolutional neural network model or a recurrent neural network model.

[0085] Step 6: Use the sample data set constructed in step 4 to train the groove size prediction model constructed in step 5 to obtain the target prediction model.

[0086] In the training process of the groove size prediction model, the Adam optimizer and the automatic parameter tuning algorithm based on Bayesian optimization are used to optimize the model parameters.

[0087] Step 7: Obtain the real-time process parameters of the aluminum electrolytic cell to be predicted, and use the target prediction model obtained in step 6 to predict the real-time process parameters to obtain the predicted values ​​of the cell wall thickness and the cell shell temperature.

[0088] Step 8: Obtain the measured value of the cell wall thickness of a certain anode of the aluminum electrolysis cell to be predicted.

[0089] The anode of an aluminum electrolytic cell has a limited service life, which is determined by the amount of anode residue and usage. When the service life is reached, the anode needs to be replaced. When the anode is replaced manually, the cell wall thickness can be measured, that is, the cell wall thickness measurement value can be obtained, so as to judge whether the target prediction model is suitable for the current operating conditions of the aluminum electrolytic cell to be predicted based on the cell wall thickness measurement value and the cell wall thickness prediction value.

[0090] Step 9: Determine whether the target prediction model meets the prediction accuracy requirements based on the groove side thickness measurement value obtained in step 8 and the groove side thickness prediction value obtained in step 7; if so, output the groove side thickness prediction value and the groove shell temperature prediction value under the real-time process parameters; if not, adjust the physical performance parameters of the aluminum electrolytic cell multi-physical field simulation model, and based on the adjusted aluminum electrolytic cell multi-physical field simulation model, repeat steps 3, 4, 6 (retrain the target prediction model), 7 and 9 until the target prediction model meets the prediction accuracy requirements, indicating that the target prediction model can adapt to the current working conditions.

[0091] When the target prediction model cannot meet the prediction accuracy requirements, the physical performance parameters of the multi-physical field simulation model of the aluminum electrolytic cell are adjusted by comprehensively considering the effects of heat transfer on the side of the electrolytic cell, material performance attenuation and operating condition changes on the cell wall thickness. The cell wall thickness and cell shell temperature under different process parameters are obtained again based on the adjusted multi-physical field simulation model of the aluminum electrolytic cell, and then a sample data set is constructed again. The target prediction model is retrained using the sample data set, and then the retrained target prediction model is used to predict the cell wall thickness and cell shell temperature, ensuring that the target prediction model can adapt to different operating conditions.

[0092] In a specific embodiment of the present invention, the present invention also develops an open API interface to support the real-time integration of the target prediction model and the industrial control system, builds a real-time feedback closed loop between the dynamic prediction and the industrial control system, dynamically optimizes the target prediction model based on the tank side dynamic prediction data and tank side measurement data, expands the application scope of the target prediction model, and realizes cross-specification and cross-operating condition prediction. By introducing an open API interface, the target prediction model is supported to quickly adapt to different aluminum electrolytic cell specifications and complex operating conditions, thereby improving the versatility and flexibility of the prediction model.

[0093] The invention is applied to a 420kA aluminum electrolytic cell, and can predict the thickness of the cell wall and the temperature change of the cell shell in real time, with the prediction error of the cell wall thickness less than 3% and the prediction accuracy of the temperature change trend of the cell shell reaching more than 95%. Table 3 shows the prediction results of anodes with different numbers.

[0094] Table 3 Prediction results corresponding to different anode numbers

[0095]

[0096]

[0097] The coordinate positions in Table 3 are based on a three-dimensional coordinate system, which takes the center line of the flue end of the aluminum electrolytic cell, the starting point at the junction of the cell shell and the bottom lining as the coordinate origin, the upward direction as the X-axis, the left and right direction as the Y-axis, and the direction toward the aluminum outlet as the Z-axis.

[0098] Figure 3 A visualization schematic diagram of the dynamic prediction results of the cell side is shown. According to the dynamic prediction method of the present invention, the real-time cell side thickness of the aluminum electrolytic cell is obtained and visualized, wherein the color band represents the cell side thickness, and the sub-image is the cell side interface image at the corresponding position. The prediction time of the present invention is 1 minute, and the prediction accuracy is 95.31% of the offline prediction, but the offline prediction time is 10 to 20 hours.

[0099] Traditional groove calculation is usually based on fixed physical equations, empirical formulas or simple neural network algorithms, which do not fully consider the dynamic changes of material properties, resulting in a significant decrease in calculation accuracy under non-standard working conditions. The present invention integrates structural parameters, process parameters and multi-scale simulation methods, and combines deep learning technology to dynamically adjust the parameters of the prediction model, thereby improving the adaptability to complex working conditions and prediction accuracy.

[0100] Compared with traditional methods, which are usually based on one-time model construction and lack of dynamic iterative optimization mechanism, resulting in poor accuracy and robustness of the prediction model, and the parameters are designed for specific electrolytic cells and are difficult to apply under different specifications and working conditions, the present invention evaluates the adaptability of the target prediction model under the current working conditions based on the trough side thickness measured when the anode is replaced and the predicted value of the trough side thickness, adjusts the multi-physical field simulation model of the aluminum electrolytic cell when the prediction accuracy requirement is not met, and retrains the target prediction model to ensure that the prediction results are highly consistent with the actual working conditions.

[0101] Example 2

[0102] An embodiment of the present invention also provides an electronic device, which includes: a memory, a processor, and a computer program / instructions stored in the memory, and the processor executes the computer program / instructions to implement the aluminum electrolysis cell slab dynamic prediction method in the embodiment of the present application.

[0103] Although not shown, the electronic device includes a processor, which can perform various appropriate operations and processes according to the program and / or data stored in the read-only memory (ROM) or the program and / or data loaded from the storage portion into the random access memory (RAM). The processor can be a multi-core processor, or it can include multiple processors. In some embodiments, the processor can include a general main processor and one or more special coprocessors, such as a central processing unit, a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. In RAM, various programs and data required for device operation are also stored. The processor, ROM and RAM are connected to each other via a bus. The input / output (I / O) interface is also connected to the bus.

[0104] The processor and the memory are used together to execute the program / instructions stored in the memory. When the program / instructions are executed by the computer, the methods, steps or functions described in the above embodiments can be implemented.

[0105] Although not shown, an embodiment of the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the method for dynamically predicting the sidewall of an aluminum electrolysis cell in an embodiment of the present application.

[0106] Readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk-read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0107] Although not shown, an embodiment of the present invention further provides a computer program product, including: a computer program / instruction, which, when executed by a processor, implements the method for dynamically predicting the aluminum electrolysis cell sidewall in the embodiment of the present application.

[0108] What is disclosed above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or modifications within the technical scope disclosed in the present invention, which should be covered within the protection scope of the present invention.

Claims

1. A method for dynamically predicting the sidewall of an aluminum electrolysis cell, characterized in that: The prediction method comprises: Acquire structural parameters and process parameters of the aluminum electrolysis cell, and construct a multi-physical field simulation model of the aluminum electrolysis cell according to the structural parameters and process parameters; Based on the multi-physics field simulation model of the aluminum electrolytic cell, the cell wall thickness and cell shell temperature under different process parameters are obtained; Construct a sample data set based on the groove side thickness and groove shell temperature under different process parameters; Constructing a groove side dimension prediction model, and using the sample data set to train the groove side dimension prediction model to obtain a target prediction model; Acquire the real-time process parameters of the aluminum electrolytic cell to be predicted, and use the target prediction model to predict the real-time process parameters to obtain a predicted value of the cell side thickness and a predicted value of the cell shell temperature; Obtaining a measured value of the thickness of a certain anode of an aluminum electrolytic cell to be predicted; According to the groove side thickness measurement value and the groove side thickness prediction value, it is judged whether the target prediction model meets the prediction accuracy requirement; if so, the groove side thickness prediction value and the groove shell temperature prediction value under the real-time process parameters are output; if not, the physical performance parameters of the aluminum electrolytic cell multi-physical field simulation model are adjusted, and based on the adjusted aluminum electrolytic cell multi-physical field simulation model, the steps of constructing the sample data set, retraining the target prediction model, predicting the real-time process parameters and judging the prediction accuracy of the target prediction model are repeated.

2. The method for dynamic prediction of the aluminum electrolysis cell sidewall according to claim 1, characterized in that: The structural parameters include geometric dimensions and physical performance parameters. The physical performance parameters include the electrical conductivity of the materials in each part and the thermal conductivity under different temperatures and different service life stage working conditions. The process parameters include temperature field, current distribution, aluminum level and electrolyte level.

3. The method for dynamic prediction of the aluminum electrolysis cell sidewall according to claim 2, characterized in that: The nanoCT method is used to obtain the electrical conductivity and thermal conductivity of the material. The specific implementation process is as follows: Avizo software was used to process the acquired CT material data to obtain a three-dimensional sampling model; Repairing the three-dimensional sampling model in the model processing software to obtain a repaired model; Importing the repaired model into a mesh drawing software, and meshing the repaired model to obtain a material mesoscopic model; The material mesoscopic model is imported into finite element simulation software, and the temperature boundary conditions and thermal conductivity boundary conditions are loaded into the material mesoscopic model to calculate the comprehensive electrical conductivity and thermal conductivity of the material.

4. The method for dynamic prediction of the aluminum electrolysis cell sidewall according to claim 1, characterized in that: Based on the multi-physics field simulation model of the aluminum electrolytic cell, the cell wall thickness and cell shell temperature under different process parameters are obtained, specifically including: The process parameters, physical boundary conditions and parameter constraints are loaded on the multi-physical field simulation model of the aluminum electrolysis cell, and an iterative solution is performed to obtain the cell wall thickness and cell shell temperature under the process parameters.

5. The method for dynamic prediction of the aluminum electrolysis cell sidewall according to claim 4, characterized in that: The physical boundary conditions and parameter constraints are optimized using a multi-scale simulation method.

6. The method for dynamic prediction of the aluminum electrolysis cell sidewall according to claim 1, characterized in that: A sample data set is constructed based on the groove side thickness and groove shell temperature under different process parameters, including: Performing denoising and missing fill processing on the process parameters; Standardize and normalize the processed process parameters; Integration and format conversion of normalized process parameters; Statistical analysis and machine learning techniques are used to extract key features from the process parameters after format conversion, and deep learning algorithms are used to evaluate the correlation between each key feature and the groove side thickness and groove shell temperature. The relevant features are determined according to the evaluation results, and a sample data set is constructed based on the determined relevant features and the corresponding groove side thickness and groove shell temperature.

7. The method for dynamic prediction of the aluminum electrolysis cell sidewall according to any one of claims 1 to 6, characterized in that: During the training process of the groove side dimension prediction model or the target prediction model, an automatic parameter tuning algorithm based on Bayesian optimization is used to optimize model parameters.

8. An electronic device comprising a memory, a processor, and a computer program / instruction stored in the memory, characterized in that: The processor executes the computer program / instructions to implement the method for dynamically predicting the ridge of an aluminum electrolysis cell according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the method for dynamically predicting the ridge of an aluminum electrolysis cell as claimed in any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the method for dynamically predicting the ridge of an aluminum electrolysis cell as claimed in any one of claims 1 to 7 is implemented.

Images