Aluminum-based material intelligent design method and equipment based on 5G communication and machine learning

Abstract

The invention relates to the field of material design, and discloses an aluminum-based material intelligent design method and equipment based on 5G communication and machine learning, and the method comprises the steps: receiving collected data, such as locally preprocessed components and technological parameters, in real time through an edge computing system and a 5G mobile network base station of an aluminum-based material production line; components and process parameters are obtained through a 5G mobile network base station, phase diagram and thermodynamic property calculation is conducted through a thermodynamic database by means of a phase diagram calculation method, and a comprehensive database is established; inputting the components and the process parameters into a plurality of performance index independent prediction models based on a comprehensive database and machine learning; when the predicted value is abnormal, a comprehensive performance prediction model and a non-dominated sorting genetic algorithm are adopted to carry out collaborative optimization on the multiple performance indexes and the comprehensive performance to obtain a Pareto optimal design scheme, and the Pareto optimal design scheme is issued to a production line for trial production and is fed back to a comprehensive database to enable the model to learn and update online. 5G, phase diagram calculation and machine learning are fused, and the design efficiency of the aluminum-based material is improved.

Description

Technical Field

[0001] This application relates to the field of aluminum-based material design, and in particular to an intelligent design method and device for aluminum-based materials based on 5G communication and machine learning. Background Technology

[0002] Aluminum-based materials are a class of high-performance metallic materials formed by alloying or combining aluminum with other materials, primarily including wrought aluminum alloys, cast aluminum alloys, and aluminum-based composite materials. They possess advantages such as low density, high specific strength, good corrosion resistance, excellent electrical and thermal conductivity, and recyclability, and have been widely used in aerospace, rail transportation, new energy vehicles, electronic information, construction, and marine engineering. For example, in new energy vehicles, aluminum foil is commonly used as the positive electrode current collector in power batteries to simultaneously meet the requirements of lightweight, high conductivity, and consistent processing.

[0003] The overall performance of aluminum-based materials directly affects their reliability in high-end equipment and devices. Defects such as excessive pinhole rate, thickness deviation, or poor plate shape will severely impact their performance in high-precision and high-consistency applications. Limited by the complexity of their composition and the strong interactions between elements, coupled with the high coupling of process parameters, traditional aluminum-based materials still face performance bottlenecks under extreme service conditions. For example, it is difficult to balance high strength and high ductility; lightweight designs often come with reduced corrosion resistance or conductivity; and in precision applications such as battery aluminum foil, pinhole defects, thickness uniformity, and surface quality remain key issues restricting performance improvement.

[0004] Traditional aluminum-based material design relies on experience and repeated experiments, resulting in long development cycles, high costs, and difficulty in simultaneously achieving multiple performance indicators. In recent years, materials genome engineering has driven the combination of phase diagram calculations (CALPHAD) thermodynamic calculations and artificial intelligence predictions, enabling rapid screening and optimization of composition and processes. However, existing methods are mostly based on offline data, lacking the ability to collect and feedback real-time, multi-dimensional data from the production process, leading to low efficiency in aluminum-based material design. Furthermore, optimizing for a single performance characteristic during the design process makes it difficult to balance the overall performance of aluminum-based materials, further contributing to the low design efficiency.

[0005] Therefore, improving the design efficiency of aluminum-based materials has become an important issue that urgently needs to be addressed. Summary of the Invention

[0006] The purpose of this application is to provide at least one intelligent design method and device for aluminum-based materials based on 5G communication and machine learning, which can at least solve the problem of how to improve the design efficiency of aluminum-based materials and at least achieve the effect of improving the design efficiency of aluminum-based materials.

[0007] In a first aspect, this application provides a smart design method for aluminum-based materials based on 5G communication and machine learning, including:

[0008] Based on the edge computing system deployed on the aluminum-based material production line, the system receives real-time data from the production line via a 5G mobile network base station deployed on the production line. The collected data is preprocessed locally by the preprocessing module of the edge computing system and includes composition and process parameters.

[0009] The received components and process parameters are respectively input into multiple independent performance index prediction models to obtain predicted values ​​for multiple performance indicators. The independent performance index prediction models are obtained by processing component samples, process parameter samples, and performance index samples in a comprehensive database using machine learning. The comprehensive database is constructed as follows: a first dataset is obtained from the production line via the 5G mobile network base station; phase diagram calculations and thermodynamic property calculations are performed based on a thermodynamic database using a phase diagram calculation method; candidate components are obtained based on the calculation results; the candidate components are experimentally verified to obtain component samples; a second dataset is obtained based on the process parameter samples and multiple performance index samples corresponding to the component samples; and the comprehensive database is constructed based on the first dataset and the second dataset.

[0010] If the error between the predicted value and the corresponding target value of at least one of the performance indicators is greater than the error threshold, the comprehensive performance indicators and the comprehensive performance of the aluminum-based material are optimized by a comprehensive performance prediction model and a non-dominated sorting genetic algorithm, and the composition and process parameters are designed to be adjusted.

[0011] The adjusted components and process parameters are sent to the production line for trial production via the edge computing system and the 5G mobile network base station, and the trial production results are fed back to the comprehensive database so that the performance index independent prediction model can learn and update online.

[0012] Optionally, before receiving the collected data from the production line in real time, the method further includes:

[0013] For the stages on the production line where data needs to be collected, communication connections are established with sensors, analyzers, and control systems distributed at each workstation through the 5G mobile network base station to collect data in real time.

[0014] The processes for which data needs to be collected include at least smelting, casting and rolling, hot rolling, cold rolling, heat treatment, surface treatment, and testing.

[0015] Optionally, the collected data also includes sensor data characterizing the quality of the aluminum-based material on the production line, and the method further includes:

[0016] The quality indicators represented by the sensor data are compared with the corresponding data thresholds;

[0017] Based on the comparison results, determine whether the quality of the aluminum-based material meets the preset quality standard;

[0018] When the quality of the aluminum-based material fails to meet the preset quality standard, an early warning message is issued.

[0019] Optionally, the first dataset can be obtained in the following ways:

[0020] The collected data after preprocessing by the preprocessing module is used as a data sample. The preprocessing of the preprocessing module includes data cleaning, noise reduction, format standardization and feature extraction.

[0021] Based on the data sample, the component sample, the process parameter sample, and multiple performance index samples are obtained to obtain the first dataset.

[0022] Optionally, the optimization of multiple performance indicators and the overall performance of the aluminum-based material using a comprehensive performance prediction model and a non-dominated sorting genetic algorithm, and the design and adjustment of the composition and process parameters, includes:

[0023] Obtain multiple combinations of the components and process parameters that can be adjusted;

[0024] For each of the combinations, the combination is input into multiple independent prediction models of the performance indicators to obtain the predicted values ​​of the multiple performance indicators of the combination;

[0025] From all the combinations, select the combination in which the error between the predicted value of the performance index and the corresponding target value is less than or equal to the error threshold.

[0026] The comprehensive performance prediction model is constructed based on all the selected combinations, and the comprehensive performance of each combination is predicted using the comprehensive performance prediction model.

[0027] Based on the predicted values ​​of multiple performance indicators of each combination that satisfies the target performance, the non-dominated sorting genetic algorithm is used to obtain the target set of multiple performance indicators and the combination of optimized comprehensive performance.

[0028] The components and process parameters are designed and adjusted based on the target set.

[0029] Optionally, the predicted values ​​of multiple performance indicators based on the combinations of the comprehensive performance satisfying the target performance are used to obtain a target set of multiple performance indicators and the combinations of the comprehensive performance optimization through a non-dominated sorting genetic algorithm, including:

[0030] An initial population is generated based on the predicted values ​​of multiple performance indicators corresponding to all combinations that satisfy the target performance, and individuals in the initial population are based on the predicted values ​​of a set of performance indicators corresponding to the combinations.

[0031] Using the initial population as the current population, the following iterative steps are performed on the current population: With the goal of optimizing multiple performance indicators, each individual in the current population is subjected to fast non-dominated sorting and crowding calculation to obtain the non-dominated order and crowding degree of each individual; the non-dominated order and crowding degree of all individuals are compared; the optimal individual is selected based on the comparison results; if the iteration stopping condition is met, the non-dominated solution set of the Pareto front is output, the non-dominated solution set including at least one optimal solution for each performance indicator; if the iteration stopping condition is not met, a parent population is formed based on the selected optimal individual, the parent population is subjected to crossover and mutation operations to obtain a child population, the parent population and the child population are merged to obtain a new population; the new population is used as the current population, and the iterative steps are executed again.

[0032] The target set is determined based on the non-dominated solution set.

[0033] Optionally, the step of constructing the comprehensive performance prediction model based on all the selected combinations and using the comprehensive performance prediction model to predict the comprehensive performance of each combination includes:

[0034] The predicted values ​​of the m performance indicators corresponding to the n combinations are standardized to obtain a standardized data matrix; wherein the data matrix has a size of n rows × m columns, and the i-th row and j-th column is the j-th standardized performance indicator of the i-th combination; wherein i, j, m and n are positive integers;

[0035] Calculate the correlation coefficient matrix based on the data matrix;

[0036] Calculate the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix;

[0037] From all the eigenvalues ​​and their corresponding eigenvectors, select the m largest eigenvalues ​​and their corresponding eigenvectors;

[0038] Based on the eigenvectors corresponding to the m largest selected eigenvalues, equations for m principal components are obtained, wherein the equation for each principal component is a linear combination of the standardized performance indices, and the eigenvectors are the standardized weights of the performance indices.

[0039] For each principal component, the ratio of the eigenvalue corresponding to the principal component to the sum of the eigenvalues ​​corresponding to all principal components is calculated to obtain the weight ratio of the principal component;

[0040] The comprehensive performance prediction equation is obtained by summing the products of the equations of each principal component and the weight ratios of the principal components.

[0041] For each of the n combinations, the predicted values ​​of the m performance indicators corresponding to the combination are input into the comprehensive performance prediction equation to obtain the comprehensive performance of the combination.

[0042] Secondly, this application provides an intelligent design device for aluminum-based materials based on 5G communication and machine learning, comprising:

[0043] The data receiving module is used to receive data collected from the production line in real time through a 5G mobile network base station deployed on the production line, based on an edge computing system deployed on the aluminum-based material production line. The collected data is preprocessed locally by the preprocessing module of the edge computing system. The collected data includes composition and process parameters.

[0044] An independent performance prediction module is used to input the received components and process parameters into multiple independent performance index prediction models to obtain predicted values ​​for multiple performance indicators. The independent performance index prediction models are obtained by processing component samples, process parameter samples, and performance index samples in a comprehensive database using machine learning. The comprehensive database is constructed as follows: a first dataset is obtained from the production line via the 5G mobile network base station; phase diagram calculations and thermodynamic property calculations are performed based on a thermodynamic database using a phase diagram calculation method; candidate components are obtained based on the calculation results; the candidate components are experimentally verified to obtain component samples; a second dataset is obtained based on the process parameter samples and multiple performance index samples corresponding to the component samples; and the comprehensive database is constructed based on the first dataset and the second dataset.

[0045] The design adjustment module is used to optimize multiple performance indicators and the overall performance of the aluminum-based material through a comprehensive performance prediction model and a non-dominated sorting genetic algorithm when the error between the predicted value and the corresponding target value of at least one of the performance indicators is greater than the error threshold, and to design and adjust the composition and the process parameters.

[0046] The trial production feedback module is used to send the adjusted components and process parameters to the production line for trial production through the edge computing system and the 5G mobile network base station, and to feed the trial production results back to the comprehensive database so that the performance index independent prediction model can learn and update online.

[0047] Thirdly, this application provides an electronic device, comprising:

[0048] At least one processor; and,

[0049] A memory communicatively connected to the at least one processor; wherein,

[0050] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to execute the intelligent design method for aluminum-based materials based on 5G communication and machine learning as described above.

[0051] Fourthly, this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the intelligent design method for aluminum-based materials based on 5G communication and machine learning as described above.

[0052] The advantages of this application compared to the prior art are:

[0053] This application presents an intelligent design method for aluminum-based materials based on 5G communication and machine learning. It utilizes an edge computing system deployed on an aluminum-based material production line. Through a 5G mobile network base station deployed on the production line, it receives real-time data collected from the line. This data, including composition and process parameters, is preprocessed locally by the edge computing system's preprocessing module. The composition and process parameters are obtained via the 5G mobile network base station. Phase diagram calculations are performed using a thermodynamic database to calculate phase diagrams and thermodynamic properties, establishing a comprehensive database. The composition and process parameters are then input into multiple independent performance indicator prediction models based on the comprehensive database and machine learning to obtain predicted values ​​for multiple target performance indicators. When the error between the predicted and target values ​​exceeds a preset threshold, a comprehensive performance prediction model and a non-dominated sorting genetic algorithm are used to collaboratively optimize the performance indicators and overall performance, obtaining a Pareto-optimal composition-process adjustment scheme. This scheme is then distributed to the production line via the edge computing system for small-batch trial production. The trial production results are fed back to the comprehensive database to enable the independent performance prediction models to learn and update online. This method integrates 5G real-time data acquisition, phase diagram calculation physical constraints, and machine learning prediction optimization techniques, significantly improving the efficiency, intelligence, and automation level of aluminum-based material design.

[0054] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0055] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0056] Figure 1 This is a flowchart of an embodiment of the intelligent design method for aluminum-based materials based on 5G communication and machine learning provided in this application;

[0057] Figure 2 This is a schematic diagram of a Pareto chart provided in one embodiment of this application;

[0058] Figure 3 This is a comparison chart of predicted and actual values ​​of pinhole PDUs provided in another embodiment of this application;

[0059] Figure 4 This is a thermogram showing the synergistic effect of aluminum-based alloy composition and processing conditions on strength and plasticity, provided in another embodiment of this application.

[0060] Figure 5 This is a diagonal distribution diagram showing the influence of aluminum-based alloy composition and processing conditions on strength and plasticity, provided in another embodiment of this application.

[0061] Figure 6 This is another embodiment of the non-dominated sorting genetic iterative trend provided in this application;

[0062] Figure 7 This is a schematic diagram illustrating the prediction of the comprehensive performance of aluminum-based alloy materials by comparing the comprehensive performance prediction model with the independent performance index prediction model provided in another embodiment of this application.

[0063] Figure 8 This is a schematic diagram illustrating the prediction fit of a comprehensive performance prediction model provided in another embodiment of this application;

[0064] Figure 9 This is a schematic diagram illustrating the prediction fit of an independent prediction model for performance metrics provided in another embodiment of this application;

[0065] Figure 10 This is a schematic diagram of the structure of an intelligent design device for aluminum-based materials based on 5G communication and machine learning, provided in one embodiment of this application.

[0066] Figure 11 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0068] With the development of artificial intelligence and 5G technology, the design and manufacturing of aluminum-based materials has ushered in new opportunities. Artificial intelligence, through machine learning and multi-objective optimization algorithms, can quickly identify complex nonlinear relationships between key process parameters and performance, enabling efficient modeling and prediction. 5G communication, with its high bandwidth, low latency, and massive connectivity, can achieve real-time acquisition and high-speed transmission of composition, process, and related parameters on the production line, ensuring data integrity and timeliness, thereby supporting online modeling and dynamic optimization control. The combination of machine learning and 5G communication is gradually shifting the materials research and development model from the traditional "experiment-verification" to "data-driven-prediction-optimization," significantly shortening the research and development cycle, reducing costs, and improving the accuracy and reliability of material microstructure and performance control.

[0069] Based on this, the applicant proposes an intelligent design method for aluminum-based materials based on 5G communication and machine learning. Leveraging the high speed, low latency, and massive connectivity of 5G communication, real-time transmission of massive amounts of sensor data from the production line is achieved. Combined with edge computing, rapid preprocessing and prediction are possible. Addressing the trade-offs between performance indicators, a non-dominated sorting genetic algorithm (NSGA-II) is employed for collaborative optimization of these indicators, constructing a comprehensive prediction model for multiple performance indicators and outputting a Pareto-optimal composition-process adjustment scheme. This intelligent design method for aluminum-based materials, integrating 5G real-time acquisition, phase diagram calculation of physical constraints, machine learning prediction, and multi-objective optimization, represents a crucial direction for improving design efficiency, reducing R&D costs, and achieving automation. It enables collaborative optimization of composition design, process control, and microstructure regulation, and fully leverages the advantages of machine learning and 5G communication for efficient modeling and prediction of complex parameter spaces.

[0070] The following is a detailed description of the implementation details of the intelligent design method for aluminum-based materials based on 5G communication and machine learning in this embodiment. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.

[0071] The intelligent design method for aluminum-based materials based on 5G communication and machine learning in this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities. Exemplary examples include mobile phones, tablets, wearable devices, in-vehicle devices, augmented reality (AR) / virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). It should be noted that this application does not limit the specific type of electronic device. The specific process can be as follows: Figure 1 As shown, it includes:

[0072] Step S101: Based on the edge computing system deployed on the aluminum-based material production line, the 5G mobile network base station deployed on the production line receives the collected data from the production line in real time. The collected data is preprocessed locally by the preprocessing module of the edge computing system. The collected data includes composition and process parameters.

[0073] Aluminum-based materials refer to the aluminum-based materials currently required for production on the production line, such as the aluminum foil used in the positive electrode current collector of batteries. In addition, aluminum-based materials can also be aluminum plates, strips, rods, tubes, profiles, castings, granules, and so on. From a compositional perspective, aluminum-based materials can be, but are not limited to, aluminum alloys and aluminum-based composite materials. Aluminum alloys can include, but are not limited to, copper-aluminum alloys, aluminum-silicon alloys, and aluminum-manganese alloys. Aluminum-based composite materials can include, but are not limited to, fiber-reinforced aluminum-based composites, particle-reinforced aluminum-based composites, and whisker-reinforced aluminum-based composites. Comprehensive performance indicators must include at least mechanical properties and defect information. In some examples, mechanical properties include surface hardness, friction and wear, corrosion resistance, roughness, and toughness; defect information includes strip breakage rate and pinhole number; process parameters include total rolling force, coiling tension, uncoiling tension, rolling speed, rolling oil temperature, rolling oil flow rate, rolling oil pressure, and rolling pressure. Multiple performance indicators are required, such as surface hardness, friction and wear, corrosion resistance, roughness and toughness, as well as the requirements for strip breakage rate, number of pinholes and rolling process parameters.

[0074] The production line includes multiple processes, such as casting, cold rolling, annealing, foil rolling, etc.

[0075] The general manufacturing process for aluminum-based materials can be roughly divided into the following four stages:

[0076] 1. Billet Preparation Stage: Liquid aluminum alloy is converted into solid billets (rolls or slabs) suitable for subsequent cold working. Depending on the alloy type, product requirements, and equipment conditions, continuous casting / rolling or semi-continuous casting and hot rolling are selected.

[0077] 2. Cold rolling and intermediate heat treatment stage: Through the combination of cold working and heat treatment, the material reaches the target thickness and its work hardening and recrystallization process is controlled. This process mainly includes cold rolling, intermediate heat treatment (which may include recovery annealing, recrystallization annealing or solution treatment), rolling system design (such as reduction ratio distribution, pass planning, etc.), and some auxiliary processes (edge ​​trimming, cleaning, straightening, etc.).

[0078] 3. Finishing and Finishing Stage: The final process to obtain the final product specifications and performance, which varies greatly depending on the product's application. It mainly includes precision rolling / foil rolling, heat treatment, surface treatment, and dimensional finishing.

[0079] 4. Inspection and Packaging: Inspect the finished products for dimensional tolerances, surface quality, mechanical properties (tensile strength, elongation, etc.), process performance, and specific physicochemical indicators. After passing the inspection, the finished products are packaged and put into storage.

[0080] 5G communication features high bandwidth and low latency, making it suitable for the digital transformation of production lines.

[0081] The frequency bands of 5G mobile network base stations can be set according to regional requirements. For enterprises, this can be a dedicated spectrum that can be applied for to ensure the stability and security of communication. In this way, using a private network is more flexible and can ensure the stability of critical production processes.

[0082] The type of 5G mobile network base station can be determined based on network requirements such as coverage area, signal strength needs, network bandwidth, and latency. For example, workshops may have metal interference or buildings, requiring specialized signal coverage. In most industrial scenarios, small base stations can be used to cover specific areas, ensuring low latency and high bandwidth communication. Of course, macro base stations can also be chosen for 5G mobile network base stations.

[0083] The network coverage area of ​​a 5G mobile network base station can be determined based on the network requirements of different areas of the production line. For example, the network coverage area may include warehouses, office areas, and production workshops where equipment for various processes such as casting, cold rolling, annealing, and foil rolling is located. The network bandwidth and latency requirements of the 5G mobile network base station can be configured according to the network requirements of each area. The corresponding antenna layout of the 5G mobile network base station can be selected based on the size of the workshop and the distribution of equipment to ensure uniform signal coverage.

[0084] The corresponding routers for 5G mobile network base stations can also be deployed according to actual needs.

[0085] A well-designed network topology for 5G mobile communication networks can ensure efficient signal coverage, low latency, and high reliability on the production line side. Examples of network topologies include distributed, centralized, and hybrid structures.

[0086] The method described in this embodiment can be applied to edge computing systems deployed on the production line side. Edge computing systems deployed on the production line side can process data locally as much as possible, reducing latency.

[0087] The construction of such infrastructure, such as 5G mobile network base stations, can provide core support for realizing the digital transmission and information transformation of production lines and intelligent production.

[0088] For example, before receiving the data collected from the production line in real time, the process also includes: establishing communication connections between the data collection links on the production line and the sensors, analyzers and control systems distributed at each workstation through a 5G mobile network base station to collect data in real time; wherein, the data collection links include at least smelting, casting and rolling, hot rolling, cold rolling, heat treatment, surface treatment and testing.

[0089] Process parameters and composition data from equipment at each stage of the production line can be recorded and transmitted in real time to a cloud data platform for storage. Edge computing systems can then be used to bring the cloud data platform and applications down to the production and testing equipment at each stage, enabling local data processing and analysis, reducing lead times, and improving the response speed and reliability of aluminum-based material production lines. This achieves real-time collection, transmission, storage, processing, and analysis of process parameters and product information data during the production process.

[0090] In this step, the deployment of 5G mobile network base stations ensures high-speed data transmission and real-time monitoring across all stages of the production line. The low latency and high bandwidth of 5G networks improve information transmission efficiency at each stage of production, laying the foundation for digital transformation. Supported by 5G networks, real-time data exchange and remote monitoring between equipment on the production line are possible, providing a robust network guarantee for upgrading to intelligent production lines. Based on this, the digital transformation of the production line was completed, encompassing the information technology transformation of equipment in various processes such as casting and rolling, cold rolling, annealing, and foil rolling. By digitally transforming the equipment in each process, data interconnection between devices was achieved, breaking down the limitations of information silos in traditional production lines. Equipment in each process can collect and transmit data in real time, ensuring precise parameter control and comprehensive operation records at every stage of production, providing reliable data support for subsequent production optimization and quality traceability.

[0091] For example, the method of obtaining the first dataset includes: using the collected data preprocessed by the preprocessing module as data samples, the preprocessing of the preprocessing module including data cleaning, noise reduction, format standardization and feature extraction; and obtaining component samples, process parameter samples and multiple performance index samples based on the data samples to obtain the first dataset.

[0092] The deployed edge computing system pushes cloud data platforms and applications down to production and testing equipment at each stage of the process. The edge computing nodes of this system have preprocessing modules. Data is transmitted to these modules via the low-latency, high-bandwidth 5G network, where it undergoes local cleaning, noise reduction, format standardization, and feature extraction. This reduces subsequent computational load and ensures that subsequent CALPHAD calculations, independent performance indicator predictions, comprehensive performance predictions, and optimization modules receive high-quality, structured data. By processing data at the edge computing nodes, the latency issues caused by large-scale data transmission to the cloud are avoided, improving real-time production responsiveness. Edge computing not only monitors and adjusts the operating status of equipment at each stage in real time but also predicts equipment failures through data analysis, enabling proactive maintenance and adjustments, further improving production efficiency and equipment utilization. The entire production process achieves closed-loop data management, with real-time data support and monitoring at every stage from raw material input to final product delivery, ensuring high efficiency, flexibility, and safety in production. Simultaneously, based on the data analysis capabilities of the edge computing system, the intelligence level of the production line is greatly enhanced, production scheduling is more precise, and product quality can be continuously optimized.

[0093] Step S102: Input the received components and process parameters into multiple independent prediction models for performance indicators to obtain predicted values ​​for multiple performance indicators; wherein, the independent prediction models for performance indicators are obtained by processing component samples, process parameter samples and performance indicator samples in the comprehensive database based on machine learning; the comprehensive database is constructed in the following way: a first dataset is obtained from the production line through a 5G mobile network base station, and phase diagram and thermodynamic property calculations are performed based on the thermodynamic database using the phase diagram calculation method. Candidate components are obtained based on the calculation results, and component samples are obtained by experimental verification of the candidate components. A second dataset is obtained based on the process parameter samples and multiple performance indicator samples corresponding to the component samples. The comprehensive database is constructed based on the first dataset and the second dataset.

[0094] These performance indicators can include microstructure, structure, defects, strengthening phases, impurity phases, matrix phases, and mechanical properties, among others. For example, defects can include the number of pinholes and the strip breakage rate. Mechanical properties can include surface hardness, friction and wear, corrosivity, roughness, and toughness, while process parameters include total rolling force, coiling tension, uncoiling tension, rolling speed, rolling oil temperature, rolling oil flow rate, rolling oil pressure, and rolling pressure, among others.

[0095] Leveraging the advantages of 5G communication, component samples, process parameter samples, and multiple performance index samples can be quickly obtained from the production line via 5G mobile network base stations, resulting in the initial dataset. This allows for rapid enrichment of the database.

[0096] Optionally, for CALPHAD calculations, phase diagrams and thermodynamic property calculations are performed based on a thermodynamic database to construct a thermodynamic database of the target alloy composition system required for aluminum-based materials, resulting in a second dataset. Then, a comprehensive database is established based on the acquisition of the first and second datasets.

[0097] The composition of aluminum-based materials can include aluminum and trace elements. 1060 and 1100 grade aluminum are industrially pure aluminum, but their aluminum content cannot meet usage requirements, and the impurities and harmful substances they contain will adversely affect the rolling and processing of aluminum-based materials. Therefore, it is necessary to improve the purity of aluminum. The large number of main elements in aluminum-based materials provides a wide degree of freedom in their design, but also brings significant challenges. Related technologies mainly employ a "trial and error" approach. However, this trial-and-error method is highly unreliable, leading to longer development cycles, low efficiency, and significant expenditure of human and material resources. Accurately designing the composition of aluminum-based materials is quite difficult. Therefore, phase diagram calculations and experimental verification methods can be used to design the composition of aluminum-based materials and optimize the trace element composition in the cast and rolled billets. Correspondingly, the construction of a comprehensive database can include:

[0098] Step 1: Perform phase diagram and thermodynamic property calculations based on the thermodynamic database.

[0099] Thermodynamic databases are supplemented with information on ternary alloy systems. For example, combining literature data and experimental data (such as experimental measurements), and focusing on important ternary alloy systems in the Al-Si-Mg-Fe-Cu-Ti-Zn-Mn-Ni-Re (Re = Sc, Er, La, Ce, La, etc.) multi-element alloy system, where experimental reports are scarce or information is limited, isothermal cross-sectional phase diagram analysis is conducted. First, phase equilibrium analyses are performed on ternary, quaternary, and pentagonal systems to obtain the solubility and partition coefficients of each element in the target phase. Thermodynamic calculations are then performed using the CALPHAD method to construct a phase diagram thermodynamic database for the aluminum alloy system. To supplement ternary alloy systems lacking experimental data, advanced technologies and detection methods such as alloy melting, vacuum heat treatment, X-ray diffraction, electron probe microanalysis, differential thermal analysis, scanning electron microscopy, and transmission electron microscopy can be used for experimental analysis. Isothermal phase diagrams can be plotted for key temperature points of each ternary alloy system within the range of 250-650℃. Based on experimental data and phase diagram information, thermodynamic calculations are performed using phase diagrams, and thermodynamic parameters of each phase in the alloy system are optimized through thermodynamic models, thereby establishing a thermodynamic phase diagram database for the Al-Si-Mg-Fe-Cu-Ti-Zn-Mn-Ni-Re alloy system.

[0100] For example, using Thermo-Calc software and an aluminum alloy thermodynamic database, isothermal interface phase diagrams, longitudinal section phase diagrams, and alloy thermodynamic properties are calculated at different temperatures, compositions, and pressures. The isothermal interface phase diagram is used to obtain the solubility and partition coefficient of each element in the target phase. Thus, the stable phase region distribution of each component under different temperature and pressure conditions can be predicted.

[0101] Step 2: Obtain candidate components based on the calculation results.

[0102] For example, based on the calculation results, the suitable composition and temperature range for aluminum-based materials are determined. Further, the phase fraction, volume fraction, alloy melting point, solubility line, and their relationship with temperature are calculated, and equilibrium and non-equilibrium solidification processes are analyzed. Based on the calculation and analysis results, suitable compositions for aluminum-based materials are selected as candidate compositions.

[0103] Step 3: Conduct experimental verification of the candidate components to obtain component samples.

[0104] For example, candidate compositions were verified through alloy smelting experiments, and the microstructure was characterized. Multi-scale characterization was performed on the microstructure of aluminum-based materials with different Al / Cu ratios and Mn, Zn, Sc, and Zr contents. The variation patterns of the second phase and texture under different component ratios were analyzed to determine the optimal trace element composition ratio. The influence of different atomic clusters on the second phase in the optimal alloy composition was analyzed using three-dimensional atom probe microanalysis, revealing the mechanism of action of Mn, Zn, Sc, and Zr composite modification on the Al-containing phase. Mechanical property testing was conducted using a universal testing machine to analyze the intrinsic correlation between composition, microstructure, and mechanical properties, obtaining the optimal composition that meets the requirements of aluminum-based materials and yielding composition samples.

[0105] For example, metallographic microscopy, scanning electron microscopy, X-ray diffraction and other characterization methods can be used to characterize microstructures.

[0106] Step 4: Construct the second dataset mentioned above based on the process parameter samples corresponding to the component samples and multiple performance index samples. Details are as follows:

[0107] By following steps one through three, a first data chain can be constructed, containing temperature samples, composition samples, and phase composition. This first data chain aggregates phase and performance data under different compositions and experimental conditions.

[0108] Based on this, high-throughput diffusion multi-node experiments can be conducted on important aluminum alloy systems to design and prepare alloys with varying compositional gradients. Electron probe microanalysis can then be used to observe the microstructure and perform microhardness analysis at different locations of several diffusion nodes, obtaining a second data chain containing compositional samples, microstructure samples, and hardness samples. Furthermore, by analyzing the microstructure evolution and microscopic and macroscopic properties under different processing conditions, a third data chain is obtained, containing compositional samples, process parameter samples, microstructure samples, structural samples, impurity phase samples, matrix phase samples, strengthening phase samples, defect samples, and mechanical property samples. Combining the first, second, and third data chains yields a fourth data chain containing compositional samples, process parameter samples, microstructure samples, structural samples, defect samples, strengthening phase samples, matrix phase samples, impurity phase samples, and mechanical property samples.

[0109] In practical applications, the microstructure of aluminum-based materials can be predicted based on the target performance of the comprehensive properties of the materials. Advanced characterization methods can then be used to characterize the microstructure of each process step. The composition, microstructure, and phase composition of each process step are tested and analyzed. The influence of each process step on the composition, microstructure, and phase composition is analyzed, as well as the influence of each process step on impurities, inclusions, and oxide phases.

[0110] The detection of surface information of aluminum-based materials can include the detection of pinholes and defects using online surface defect detection systems and machine vision technology. Deep learning neural network algorithms are used to identify and quantitatively analyze information on pinholes, defects, and broken bands, and the detected pinhole data, defect data, and broken band rates are stored.

[0111] The prediction and characterization analysis of data related to aluminum-based materials includes: predicting the microstructure of aluminum-based materials, including phase composition, volume fraction of each phase, grain size, texture, etc., as well as the microstructure, volume fraction, and crystal structure of possible impurity phases. Through screening, aluminum-based materials with no or few impurity phases are obtained. Advanced characterization techniques such as transmission electron microscopy, scanning electron microscopy, electron probe microanalysis, backscattered electron diffraction, differential thermal analysis, and X-ray diffraction are used to characterize the microstructure of aluminum-based materials in each process. The composition, microstructure, and phase composition of each process are tested and analyzed, and the influence of different processes on the composition, microstructure, and phase composition are analyzed, especially the effects on impurity phases, inclusions, and oxide phases.

[0112] Based on the fourth data chain, a second dataset is constructed. Thus, a comprehensive and rich database can be built by merging the second dataset, the first dataset, laboratory data, and data obtained from published literature.

[0113] In practical applications, for each performance indicator, an independent prediction model for that performance indicator can be pre-built. The independent prediction model for the performance indicator can be obtained by processing the database using techniques such as ridge regression, least squares method, decision tree, support vector machine, random forest, boosting tree, extreme gradient boosting, artificial neural network and K-nearest neighbor algorithm.

[0114] Before building an independent predictive model for performance metrics, the database must be preprocessed. Preprocessing includes data cleaning and data transformation. Data cleaning involves handling missing values, outliers, and duplicates. When handling missing values, imputation methods such as using the mean, median, or mode can be used, or rows or columns containing missing values ​​can be deleted. When handling outliers, statistical methods such as box plots and Z-scores can be used to detect and process them, ensuring data quality and accuracy. When handling duplicates, they can be checked and deleted to prevent them from affecting model training. Data transformation includes standardization and normalization. Standardization transforms the data into a distribution with a mean of 0 and a standard deviation of 1, ensuring that all data have the same dimensions, facilitating model training. Normalization maps the data to the [0,1] interval, which can be used in distance-sensitive algorithms such as K-nearest neighbors and support vector machines, allowing all features to be compared on the same scale. When using artificial neural networks, data splitting is also necessary. Data splitting involves dividing the data into training and test sets. The training set is used for model training and parameter optimization, while the test set is used to evaluate the model's performance on unseen data. The comprehensive data samples in the database are split according to a certain ratio (e.g., 70% training set, 30% test set, or 80% training set, 20% test set). For example, the `train_test_split` function can be used to divide the data into training and test sets.

[0115] The loss function used during training can be the mean squared error (MSE). MSE is suitable for regression problems. Use MSE and the coefficient of determination to evaluate the model's performance on the test set and examine its predictive power.

[0116] Cross-validation (such as K-fold cross-validation) can be used to further verify the model's generalization ability and improve its stability.

[0117] For example, data preprocessing can be performed using tools such as Datameer, Data Ladder, or OpenRefine.

[0118] During implementation, the database and independent performance indicator prediction models can be built through a cloud-based data platform.

[0119] Step S103: When the error between the predicted value of at least one performance index and the corresponding target value is greater than the error threshold, optimize multiple performance indices and the overall performance of aluminum-based materials through a comprehensive performance prediction model and a non-dominated sorting genetic algorithm, and design adjustment components and process parameters.

[0120] If the error between the predicted value and the corresponding target value of at least one performance index is greater than the error threshold, it is determined that there is anomaly in the composition and process parameters. Then, multiple performance indicators and the overall performance of aluminum-based materials are optimized by using a comprehensive performance prediction model and a non-dominated sorting genetic algorithm. The composition and process parameters are designed to adjust the error of multiple performance indicators to be less than or equal to the error threshold.

[0121] In practical applications, if the errors between the predicted values ​​and corresponding target values ​​of all performance indicators are less than or equal to the error threshold, the composition and process parameters are considered normal. If the error between the predicted value and the corresponding target value of at least one performance indicator is greater than the error threshold, an anomaly is identified in the composition and process parameters, and an anomaly alert can be issued. This allows production line staff to promptly monitor the anomaly. Furthermore, by optimizing multiple performance indicators and the overall performance of the aluminum-based material, the composition and process parameters can be designed and adjusted to ensure that the errors of multiple performance indicators are less than or equal to the error threshold, thereby enhancing the level of automation.

[0122] Step S104: The adjusted components and process parameters are sent to the production line for trial production through the edge computing system and the 5G mobile network base station, and the trial production results are fed back to the comprehensive database so that the performance index independent prediction model can learn and update online.

[0123] By transmitting the designed components and process parameters to the production line equipment via a 5G mobile network base station for small-batch trial production, the designed components and process parameters can be experimentally verified. The trial production results are then fed back to the comprehensive database, enabling the performance indicator independent prediction model to learn and update online. This real-time, multi-dimensional data collection and feedback allows for rapid intelligent design of aluminum-based materials, enhancing the overall intelligence level of aluminum-based material design.

[0124] The above steps S102-S104 can realize intelligent design of aluminum-based materials and monitoring of the production process, and can be packaged into production process data monitoring software.

[0125] In this embodiment, based on an edge computing system deployed on an aluminum-based material production line, real-time data is received from the production line via a 5G mobile network base station deployed on the production line. The collected data, including composition and process parameters, undergoes local preprocessing by the edge computing system's preprocessing module. The composition and process parameters are obtained through the 5G mobile network base station, and phase diagram calculations are performed using a thermodynamic database to establish a comprehensive database. The composition and process parameters are then input into multiple independent performance indicator prediction models based on the comprehensive database and machine learning to obtain predicted values ​​for multiple target performance indicators. When the error between the predicted value and the target value exceeds a preset threshold, a comprehensive performance prediction model and a non-dominated sorting genetic algorithm are used to collaboratively optimize the performance indicators and overall performance, obtaining the Pareto-optimal composition-process adjustment scheme. This scheme is then distributed to the production line via the edge computing system for small-batch trial production. The trial production results are fed back to the comprehensive database to enable the independent performance prediction models to learn and update online. This method integrates 5G real-time data acquisition, phase diagram calculation physical constraints, and machine learning prediction optimization technology to achieve real-time multi-dimensional data acquisition and feedback, significantly improving the efficiency, intelligence, and automation level of aluminum-based material design.

[0126] In addition, the solution in this embodiment also improves production efficiency, ensures stable production processes, and meets product performance standards.

[0127] Furthermore, the intelligent design method for aluminum-based materials based on 5G communication and machine learning in this embodiment can shorten the R&D cycle, effectively reduce resource waste, and lower production costs.

[0128] For aluminum-based material design, such as battery aluminum foil, a comprehensive solution can be proposed across all stages, including initial composition design, experimental research and development, pilot production, mass production, and 5G digital manufacturing. This will ultimately form a complete technical system suitable for the production of 1070 aluminum foil battery cathode materials, including process documents, operating procedures, quality control standards, and digital system management programs. This ensures production stability and consistency, guaranteeing the normal production and stable performance of 1070 aluminum foil. This technology can be applied to large-scale production, conducting industrialization demonstration projects and extending to the mass production of other aluminum-based materials. Based on industrial application demonstrations, the application effects of 5G communication in aluminum-based material production can be showcased, further accelerating the promotion and application of the technology, while also promoting its widespread application in other aluminum deep processing fields. This will accelerate the establishment of an innovative industrial cluster for new aluminum-based materials.

[0129] In some embodiments, the collected data also includes sensor data characterizing the quality of aluminum-based materials on the production line. The intelligent design method for aluminum-based materials based on 5G communication and machine learning further includes: comparing the quality indicators characterized by the sensor data with the corresponding data thresholds; determining whether the quality of the aluminum-based materials meets the preset quality standards based on the comparison results; and issuing a warning message when the quality of the aluminum-based materials does not meet the preset quality standards.

[0130] The sensor data can include image sensor data of the aluminum-based material surface. This image sensor data can be used to obtain characterizing quality indicators. These quality indicators can include defect information. Severe defects in the aluminum-based material indicate that the quality has not met standards.

[0131] In practical implementation, multi-production manufacturing system quality management software can be designed to manage the quality of multiple production lines. Specifically, online quality inspection can be performed on multiple production lines, that is, the quality indicators represented by sensor data are compared with the corresponding data thresholds; based on the comparison results, it is determined whether the quality of aluminum-based materials meets the preset quality standards; when the aluminum-based materials do not meet the preset quality standards, an early warning message is issued.

[0132] The data thresholds and preset quality standards can be set according to actual needs.

[0133] In this embodiment, online quality detection using sensor data enables precise quality control of each stage of the production process in real time, determining whether the produced aluminum-based materials meet quality standards. If the quality standards are not met, timely warning information is issued, quickly capturing potential quality deviations or defects, ensuring stable product quality, and rapidly identifying and resolving quality problems in production. This achieves efficient and accurate quality judgment, significantly improving production efficiency and product quality.

[0134] Furthermore, the results of quality assessments can be visualized. This graphical presentation helps analysts quickly identify potential quality fluctuations during production, allowing for timely adjustments to production strategies, thereby optimizing the decision-making process and improving the accuracy and efficiency of the production line.

[0135] Multi-production manufacturing system quality management software can offer modular component designs, which are used to set up quality inspection processes according to business needs. This allows for flexible adjustments as needed, ensuring adaptability to different quality management requirements.

[0136] If 5G communication is used to transmit collected data, its low latency and high bandwidth enable real-time transmission, allowing for real-time processing and accurate judgment. Once a quality issue is detected, the system can quickly issue an early warning and automatically trigger production line adjustments or alarm mechanisms, ensuring the problem is resolved immediately. This significantly improves production efficiency and prevents large-scale quality losses.

[0137] Multi-production system quality management software can also perform raw material inspection and production process control, enabling centralized management and control. Furthermore, the digital management capabilities of this software make the entire quality control system more transparent and traceable. It can also generate quality management data reports, allowing management to better understand the quality status during production and make timely improvement measures. Thus, the software not only helps improve the automation and intelligence level of the production line but also provides technical support for continuous quality optimization, ensuring the stability and consistency of product quality.

[0138] In some embodiments, the intelligent design method for aluminum-based materials based on 5G communication and machine learning may further include: inputting composition and process parameters into a production prediction model to obtain a predicted production output; wherein the production prediction model is obtained by training an initial neural network model based on composition samples, process parameter samples, and production output samples; and analyzing the production output of the production line based on the predicted production output.

[0139] In practical applications, production samples can be obtained from historical output. Then, based on these samples, along with component and process parameter samples, an initial neural network model can be trained to obtain a production prediction model. Based on the predicted output, when analyzing the production line's output, we can analyze output trends, production efficiency, and other aspects.

[0140] In this embodiment, by predicting the output through real-time collection of component and process parameters, the production process can be further monitored, making it easier for staff to adjust the production line control strategy in a timely manner.

[0141] In some embodiments, multiple performance indicators and the overall performance of aluminum-based materials are optimized using a comprehensive performance prediction model and a non-dominated sorting genetic algorithm, and adjustments to composition and process parameters are designed, including:

[0142] Obtain multiple combinations of components and process parameters that can be adjusted;

[0143] For each combination, the combination is input into multiple independent prediction models for performance indicators to obtain the predicted values ​​of multiple performance indicators for the combination.

[0144] From all combinations, select the combination in which the error between the predicted value of multiple performance indicators and the corresponding target value is less than or equal to the error threshold.

[0145] A comprehensive performance prediction model is constructed based on all selected combinations, and the comprehensive performance of each combination is predicted using the comprehensive performance prediction model.

[0146] Based on the predicted values ​​of multiple performance indicators of each combination that meets the target performance, a non-dominated sorting genetic algorithm is used to obtain the target set of multiple performance indicators and combinations with optimized overall performance.

[0147] The design adjusts the composition and process parameters based on the target set.

[0148] The target performance can be set according to the requirements of aluminum-based materials.

[0149] The target set is a collection of multiple performance metrics and combinations of comprehensive performance optimizations.

[0150] After designing and adjusting the components and process parameters based on the target set, specifically, a component and process parameter can be randomly selected from the target set and transmitted to the production line via a 5G mobile network base station, enabling the equipment on the production line to adjust the component and process parameters. For example, the equipment on the production line includes a Programmable Logic Controller (PLC). The selected component and process parameters can be transmitted to the PLC in the production line equipment via a 5G mobile network base station.

[0151] Non-dominated sorting genetic algorithm is a multi-objective optimization algorithm based on genetic algorithm.

[0152] In this embodiment, a non-dominated sorting genetic algorithm can be used to quickly optimize multiple performance indicators simultaneously, thereby optimizing the overall performance of aluminum-based materials and enabling automated and rapid adjustment of composition and process parameters.

[0153] In some embodiments, a comprehensive performance prediction model is constructed based on all selected combinations, and the comprehensive performance prediction model is used to predict the comprehensive performance of each combination, including:

[0154] Step 1: Standardize the predicted values ​​of the m performance indicators corresponding to the n combinations to obtain a standardized data matrix; the data matrix has a size of n rows × m columns, and the i-th row and j-th column is the j-th standardized performance indicator of the i-th combination; where i, j, m and n are positive integers.

[0155] In this step, the predicted values ​​of each performance index corresponding to the n combinations are dimensionless to obtain standardized performance indices:

[0156]

[0157] in, The standardized data matrix B = (b ij ) n×m ,in

[0158] Among them, y j Let x be the j-th standardized performance metric vector; j The original performance metric vector; μ j Let be the mean of the j-th performance metric; s j Let a be the standard deviation of the j-th indicator; ij Let be the j-th original performance metric for the i-th combination; in the data matrix B, n is the row of the matrix; m is the column of the matrix; b ij Let be the j-th standardized performance index of the i-th combination.

[0159] Step 2: Calculate the correlation coefficient matrix based on the data matrix.

[0160] Specifically, for data matrix B, calculate the correlation coefficient matrix R′=(r lc ) n×m .in

[0161] Among them, b il b is the standardized performance index of the l-th combination in the i-th combination; ic Let r be the standardized performance index of the c-th component in the i-th combination, where l = 1, 2, ..., m, c = 1, 2, ..., m; lc Let be the correlation coefficient value of the l-th and c-th standardized performance indicators of the i-th combination in the coefficient matrix R′. The correlation coefficient ranges from [-1, 1] and represents the strength and direction of the relationship between the two variables.

[0162] Step 3: Calculate the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix.

[0163] Specifically, the eigenvalues ​​λ1≥λ2≥…≥λ can be calculated from the correlation coefficient matrix R′. m , and the corresponding orthogonalized eigenvectors.

[0164] Step 4: From all eigenvalues ​​and their corresponding eigenvectors, select the eigenvectors corresponding to the m largest eigenvalues.

[0165] Step 5: Based on the eigenvectors corresponding to the m largest selected eigenvalues, obtain the equations of the m principal components. The equation of each principal component is a linear combination of standardized performance indices, and the eigenvectors are the standardized weights of each performance indices.

[0166] Equations with m principal components: Specifically:

[0167]

[0168] Among them, u i′j Let f(x) be the coefficient of the j-th variable in the i′-th principal component, i.e., the weight of each variable in the principal component. j ) represents the original performance index vector x j The standardized value, Y i′ Let i be the i′-th principal component, where i′ = 1, 2, 3, ..., m; j = 1, 2, 3, ..., m.

[0169] Step 6: For each principal component, calculate the ratio of the eigenvalue corresponding to the principal component to the sum of the eigenvalues ​​corresponding to all principal components to obtain the weight ratio of the principal component.

[0170] Specifically, the eigenvalues ​​corresponding to each principal component can be summed, and then the ratio of the eigenvalue of each principal component to the summation result can be calculated to obtain the weight ratio of each principal component. For example, the weight ratio of each principal component can be calculated according to the following formula.

[0171]

[0172] Where: b i′ λ represents the weight ratio of the i′-th principal component; k Let be the eigenvalue corresponding to the k′-th principal component, where k′ = 1, 2, 3, ..., m. The weight ratio reflects the contribution.

[0173] Step 7: Based on the sum of the products of the equations of each principal component and the weight ratios of the principal components, the comprehensive performance prediction equation is obtained.

[0174] For example, the comprehensive performance prediction equation is as follows:

[0175]

[0176] Among them, Y com Indicates overall performance.

[0177] Step 8: For each of the n combinations, input the predicted values ​​of the m performance indicators corresponding to the combination into the comprehensive performance prediction equation to obtain the comprehensive performance of the combination.

[0178] In this embodiment, the comprehensive performance is obtained by extracting the principal components, ensuring that the comprehensive performance prediction model can accurately capture the nonlinear relationship between the combination of components and process parameters and performance, accurately predict the comprehensive performance, and improve the comprehensive performance of aluminum-based materials in the production process.

[0179] SHAP Value is used to measure the contribution of eigenvalues ​​for data analysis. By analyzing multivariate multicollinearity, the impact of single and multiple variables on predicted values ​​is analyzed, and the synergistic effect between different variables is further calculated.

[0180] Alternatively, recursive feature elimination (RFE) or entropy methods can be used to construct a comprehensive performance prediction model.

[0181] Current independent performance prediction models are insufficient in their ability to comprehensively predict the performance of multifunctional new materials, especially due to the challenges of small sample sizes and large search spaces in black-box data, resulting in low prediction accuracy and difficulty in achieving ideal prediction results. This embodiment addresses these limitations by constructing a comprehensive performance prediction model that overcomes the limitations of traditional independent performance prediction models in calculating relationships. It can further calculate the quantitative relationships between different microstructures, structures, and properties of aluminum-based materials under different process conditions, significantly improving the model's prediction accuracy and solving the problems of effectiveness and reliability of prediction models.

[0182] In some embodiments, based on the predicted values ​​of multiple performance indicators of each combination whose overall performance satisfies the target performance, a target set of multiple performance indicators and combinations of optimized overall performance is obtained through a non-dominated sorting genetic algorithm, including:

[0183] Step 1: Generate an initial population based on the predicted values ​​of multiple performance indicators corresponding to all combinations that meet the target performance. Individuals in the initial population are based on the predicted values ​​of a set of performance indicators corresponding to the combination.

[0184] Step 2: Using the initial population as the current population, perform the following iterative steps on the current population: With the goal of achieving optimal performance metrics, perform fast non-dominated sorting and crowding calculation on each individual in the current population to obtain the non-dominated order and crowding degree for each individual; compare the non-dominated order and crowding degree of all individuals; select the optimal individual based on the comparison results; if the iteration stopping condition is met, output the non-dominated solution set of the Pareto front, which includes the optimal solution for at least one set of performance metrics; if the iteration stopping condition is not met, form a parent population based on the selected optimal individual, perform crossover and mutation operations on the parent population to obtain the offspring population, merge the parent population and the offspring population to obtain a new population; use the new population as the current population and re-execute the iterative steps.

[0185] The objective function of the performance index is the value of the performance index. The goal is to find the non-dominated solution with the objective function of each performance index being optimized.

[0186] The objective functions for each performance index are f1(y), f2(y), ..., f j (y), ..., f m (y), fj Let f represent the objective function of the j-th performance index out of m performance indices. j The solution y of (y) i″ The congestion level is calculated as follows:

[0187]

[0188] Where, CD(y) i″ ) is the solution y i″ The degree of congestion, f j (y i″+1 ) and f j (y i″ -1) are solutions to y i″ The objective values ​​(i.e., the values ​​of the objective function) of the right and left neighboring solutions on the j-th objective function. and These are the maximum and minimum values ​​of all solutions to the j-th objective function, respectively.

[0189] The optimal individual is selected based on the comparison results, specifically including: if individual y i″ The non-dominated layer is superior to individual y j′ The non-dominated layer in which the solution y is chosen. i″ The target value is Cd(y) if they have the same rank and i″ )>CD(y j′ ), choose solution y i″ The target value. rank represents the non-dominated order, characterizing the hierarchy of the dominance layer.

[0190] Step 3: Use the non-dominated solution set as the target set.

[0191] A solution in the non-dominated solution set is an optimized set of performance metrics that improves the overall performance of the combination corresponding to that solution. Therefore, we can obtain the corresponding combinations based on the solutions in the non-dominated solution set to get the target set of combinations with optimal overall performance.

[0192] The iteration stopping conditions include reaching the maximum number of iterations, or the overall performance of a solution in the non-dominated solution set meeting the target performance. Substituting a solution into the overall performance prediction equation yields the overall performance of that solution. Controlling the iteration stopping by the number of iterations allows for rapid optimization, while controlling the stopping by the overall performance index ensures that the combination of components and process parameters meets the overall performance requirements.

[0193] In this embodiment, multiple iterations can simultaneously optimize multiple performance indicators, thereby obtaining a set of combinations of components and process parameters with better overall performance indicators.

[0194] The following uses specific process parameters as an example to further illustrate the solution of this embodiment.

[0195] In practice, data on process parameters such as total rolling force, coiling tension, uncoiling tension, rolling speed, rolling oil temperature, rolling oil flow rate, and rolling oil pressure can be obtained to investigate their impact on pinholes in finished aluminum-based materials. Pareto analysis can be used to identify the most influential process parameters. For example... Figure 2 As shown in the figure, the three most influential parameters identified by the Pareto chart are: unwinding tension, winding tension, and rolling oil pressure. Figure 2 In this diagram, the rank represents the hierarchy of different process parameters, and the t-value on the vertical axis represents the significance of the process parameter. The larger the t-value, the less likely the factor effect is caused by random error, and the more significant the factor is. The upper and lower limits of the t-value can be set as needed.

[0196] Based on three parameters—uncoiling tension, winding tension, and rolling hydraulic pressure—the scheme in this embodiment is used for intelligent design of aluminum-based materials. By using Pareto analysis to quantify the contribution of the combined parameters, key parameters are obtained, and intelligent design of aluminum-based materials based on these key parameters is more efficient.

[0197] When predicting pinhole PDUs using an independent prediction model for pinhole PDU performance indicators, the comparison between predicted and actual values ​​is as follows: Figure 3 As shown in the figure, the prediction accuracy is very high, reaching 95.3%, with a correlation coefficient of 0.921. This is beneficial for improving the overall performance of the final aluminum-based material.

[0198] Taking the synergistic design of strength and plasticity in aluminum-based alloys as an example, a target alloy composition system is established, including the aluminum alloy Al-Si-Mg-Fe-Cu-Ti-Zn-Mn-Ni-Re (where Re represents other elements in the aluminum-based material composition). Based on industrial production and literature records, combined with experimental phase diagram information, the data chain in the Al-Si-Mg-Fe-Cu-Ti-Zn-Mn-Ni-Re system is further improved, and a database is established. The composition ratio is iteratively updated through comprehensive performance prediction and experimental verification.

[0199] Figure 4This is a heatmap showing the synergistic effects of aluminum-based alloy composition and processing conditions on strength and plasticity. Feature engineering reduces the dimensionality of high-dimensional data for intuitive visualization, from dark blue (negative correlation) to yellow (no correlation) and then to red (positive correlation), with darker colors indicating a larger absolute value of correlation. The graph shows that Solution temperature (+0.37) and Rolling reduction (+0.33) are positively correlated with strength and plasticity and should be prioritized as predictors. N (-0.05) and Air (+0.29), etc., have weaker correlations with strength and plasticity and can be excluded from feature selection.

[0200] Figure 5 This is a diagonal distribution plot showing the influence of aluminum-based alloy composition and processing conditions on strength and plasticity. It illustrates the effects of multi-metallic elements Al (aluminum), Co (cobalt), Ni (nickel), Cr (chromium), Mn (manganese), Cu (copper), Fe (iron), Ti (titanium), and Nb (niobium) on the target variable. A significant positive correlation was found between Fe and Ni, possibly due to Fe and Ni jointly enhancing strength in alloy design. A positive correlation was also found between Ni and Cr, possibly due to their synergistic effect with Cr and Ni in improving plasticity. Strength and plasticity, as well as Nb, generally show an upward trend with increasing Nb content. In the Fe-Ni-Cr triangular region, the scatter plot exhibits a dense "cloud-like" distribution, suggesting that all three elements jointly influence the material's strength and plasticity.

[0201] Figure 6 The results show the iterative trend of non-dominated sorting genetics. As the number of iterations increases, the fitness reaches its optimal level at 95 iterations.

[0202] Figure 7 The diagram shows the prediction of the comprehensive performance of aluminum-based alloy materials by comparing the comprehensive performance prediction model (i.e., the comprehensive performance model) with the independent performance index prediction model (i.e., the independent model). It is clear that the accuracy of the comprehensive performance prediction model is higher than that of the independent performance index prediction model.

[0203] Figure 8 and Figure 9 The diagram shows the fitting of the comprehensive performance prediction model to the independent performance index prediction model for aluminum-based alloy materials. It is clear that the comprehensive performance prediction model has a better fit than the independent performance index prediction model.

[0204] The results show that the determination coefficient R of the comprehensive performance prediction model is 2 The accuracy reached 0.9947, with a prediction accuracy rate of 98.72%.

[0205] The above-mentioned embodiments can be implemented in combination according to actual needs. It should be understood that the sequence number of each step in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0206] The above-mentioned embodiments can be implemented in combination with each other.

[0207] Another embodiment of this application relates to a smart design device for aluminum-based materials based on 5G communication. The smart design device for aluminum-based materials based on 5G communication described below can be referred to in correspondence with the smart design method for aluminum-based materials based on 5G communication described above. The implementation details of the smart design device for aluminum-based materials based on 5G communication in this embodiment are described in detail below. The following implementation details are provided only for ease of understanding and are not necessary for implementing this solution. The schematic diagram of the smart design device for aluminum-based materials based on 5G communication in this embodiment can be seen as follows: Figure 10 As shown, it includes:

[0208] The data receiving module 1001 is used to receive the collected data of the production line in real time through the 5G mobile network base station deployed on the production line, based on the edge computing system deployed on the aluminum-based material production line. The collected data is preprocessed locally by the preprocessing module of the edge computing system. The collected data includes composition and process parameters.

[0209] An independent performance prediction module 1002 is used to input the received components and process parameters into multiple independent performance index prediction models to obtain predicted values ​​for multiple performance indicators. The independent performance index prediction models are obtained by processing component samples, process parameter samples, and performance index samples in a comprehensive database using machine learning. The comprehensive database is constructed as follows: a first dataset is obtained from the production line via the 5G mobile network base station; phase diagram calculations and thermodynamic property calculations are performed based on a thermodynamic database using a phase diagram calculation method; candidate components are obtained based on the calculation results; the candidate components are experimentally verified to obtain component samples; a second dataset is obtained based on the process parameter samples and multiple performance index samples corresponding to the component samples; and the comprehensive database is constructed based on the first dataset and the second dataset.

[0210] The design adjustment module 1003 is used to optimize the comprehensive performance of multiple performance indicators and the comprehensive performance of the aluminum-based material through a comprehensive performance prediction model and a non-dominated sorting genetic algorithm when the error between the predicted value and the corresponding target value of at least one of the performance indicators is greater than the error threshold, and to design and adjust the composition and the process parameters.

[0211] The trial production feedback module 1004 is used to send the adjusted components and process parameters to the production line for trial production through the edge computing system and the 5G mobile network base station, and to feed back the trial production results to the comprehensive database so that the performance index independent prediction model can learn and update online.

[0212] In some embodiments, it further includes: a communication establishment module;

[0213] The communication establishment module is used to establish communication connections with sensors, analyzers and control systems distributed at each workstation through the 5G mobile network base station for the links on the production line where data needs to be collected, so as to collect data in real time.

[0214] The processes for which data needs to be collected include at least smelting, casting and rolling, hot rolling, cold rolling, heat treatment, surface treatment, and testing.

[0215] In some embodiments, the acquired data further includes sensor data characterizing the quality of the aluminum-based material on the production line, and the device further includes: a quality detection module, the quality detection module being used for:

[0216] The quality indicators represented by the sensor data are compared with the corresponding data thresholds;

[0217] Based on the comparison results, determine whether the quality of the aluminum-based material meets the preset quality standard;

[0218] When the quality of the aluminum-based material fails to meet the preset quality standard, an early warning message is issued.

[0219] In some embodiments, the apparatus further includes a database building module, the database building module being configured to:

[0220] The collected data after preprocessing by the preprocessing module is used as a data sample. The preprocessing of the preprocessing module includes data cleaning, noise reduction, format standardization and feature extraction.

[0221] Based on the data sample, the component sample, the process parameter sample, and multiple performance index samples are obtained to obtain the first dataset.

[0222] In some embodiments, the adjustment module 1003 is designed to be used specifically for:

[0223] Obtain multiple combinations of the components and process parameters that can be adjusted;

[0224] For each of the combinations, the combination is input into multiple independent prediction models of the performance indicators to obtain the predicted values ​​of the multiple performance indicators of the combination;

[0225] From all the combinations, select the combination in which the error between the predicted value of the performance index and the corresponding target value is less than or equal to the error threshold.

[0226] The comprehensive performance prediction model is constructed based on all the selected combinations, and the comprehensive performance of each combination is predicted using the comprehensive performance prediction model.

[0227] Based on the predicted values ​​of multiple performance indicators of each combination that satisfies the target performance, the non-dominated sorting genetic algorithm is used to obtain the target set of multiple performance indicators and the combination of optimized comprehensive performance.

[0228] The components and process parameters are designed and adjusted based on the target set.

[0229] In some embodiments, the adjustment module 1003 is designed to be used specifically for:

[0230] An initial population is generated based on the predicted values ​​of multiple performance indicators corresponding to all combinations that satisfy the target performance, and individuals in the initial population are based on the predicted values ​​of a set of performance indicators corresponding to the combinations.

[0231] Using the initial population as the current population, the following iterative steps are performed on the current population: With the goal of optimizing multiple performance indicators, each individual in the current population is subjected to fast non-dominated sorting and crowding calculation to obtain the non-dominated order and crowding degree of each individual; the non-dominated order and crowding degree of all individuals are compared; the optimal individual is selected based on the comparison results; if the iteration stopping condition is met, the non-dominated solution set of the Pareto front is output, the non-dominated solution set including at least one optimal solution for each performance indicator; if the iteration stopping condition is not met, a parent population is formed based on the selected optimal individual, the parent population is subjected to crossover and mutation operations to obtain a child population, the parent population and the child population are merged to obtain a new population; the new population is used as the current population, and the iterative steps are executed again.

[0232] The target set is determined based on the non-dominated solution set.

[0233] In some embodiments, the adjustment module 1003 is designed to be used specifically for:

[0234] The predicted values ​​of the m performance indicators corresponding to the n combinations are standardized to obtain a standardized data matrix; wherein the data matrix has a size of n rows × m columns, and the i-th row and j-th column is the j-th standardized performance indicator of the i-th combination; wherein i, j, m and n are positive integers;

[0235] Calculate the correlation coefficient matrix based on the data matrix;

[0236] Calculate the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix;

[0237] From all the eigenvalues ​​and their corresponding eigenvectors, select the m largest eigenvalues ​​and their corresponding eigenvectors;

[0238] Based on the eigenvectors corresponding to the m largest selected eigenvalues, equations for m principal components are obtained, wherein the equation for each principal component is a linear combination of the standardized performance indices, and the eigenvectors are the standardized weights of the performance indices.

[0239] For each principal component, the ratio of the eigenvalue corresponding to the principal component to the sum of the eigenvalues ​​corresponding to all principal components is calculated to obtain the weight ratio of the principal component;

[0240] The comprehensive performance prediction equation is obtained by summing the products of the equations of each principal component and the weight ratios of the principal components.

[0241] For each of the n combinations, the predicted values ​​of the m performance indicators corresponding to the combination are input into the comprehensive performance prediction equation to obtain the comprehensive performance of the combination.

[0242] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0243] Another embodiment of this application relates to an electronic device, such as... Figure 11 As shown, it includes: at least one processor 1101; and a memory 1102 communicatively connected to the at least one processor 1101; wherein the memory 1102 stores instructions executable by the at least one processor 1101, the instructions being executed by the at least one processor 1101 to enable the at least one processor 1101 to execute the intelligent design method for aluminum-based materials based on 5G communication and machine learning in the above embodiments.

[0244] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0245] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0246] Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0247] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0248] In its specific implementation, all computer programs involved in this invention are developed using the Python language within the PyCharm Community Edition software:

[0249] 1) From the development of the core program to subsequent optimization and upgrades, especially in the data reading method, the original sequential reading was changed to the sliding window averaging method instead of the original grouping method, which significantly improved the data reading speed; at the same time, the multi-performance integrated prediction model and data were migrated to the GPU processor for parallel computing, which greatly improved the program's running efficiency.

[0250] 2) The program was improved using a genetic algorithm, with the addition of an optimal individual retention mechanism, a fast training strategy for fitness evaluation, and the introduction of a visualization function for the evolutionary process.

[0251] 3) Optimize neural network training, implement an early stopping mechanism to prevent overfitting, use the Adam optimizer instead of SGD, and add validation set monitoring.

[0252] Before optimization, it took the program about 6 seconds to read 4,000 data entries; after optimization, this was reduced to 2-3 seconds. When processing 4,000 data entries, the program's running time was reduced from 100-120 seconds to 25-30 seconds, resulting in a significant improvement in overall running efficiency.

[0253] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A 5G communication and machine learning based intelligent design method for aluminum-based materials, characterized by, The method comprises the following steps: Based on the edge computing system deployed on the production line of aluminum-based materials, the acquisition data of the production line is received in real time through the 5G mobile network base station deployed on the production line, the acquisition data is preprocessed locally by the preprocessing module of the edge computing system, and the acquisition data includes composition and process parameters; The received composition and process parameters are respectively input into a plurality of performance index independent prediction models to obtain a plurality of performance index prediction values; wherein the performance index independent prediction model is obtained by processing the composition sample, process parameter sample and performance index sample in the comprehensive database based on machine learning; the comprehensive database is constructed by the following methods: a first data set obtained from the production line through the 5G mobile network base station, and a phase diagram and thermodynamic property calculation based on a thermodynamic database using a phase diagram calculation method, obtaining candidate components based on the calculation results, obtaining the composition sample by experimental verification, obtaining a second data set based on the process parameter sample and a plurality of performance index samples corresponding to the composition sample, and constructing the comprehensive database based on the first data set and the second data set; In the case that the error between the prediction value of at least one performance index and the corresponding target value is greater than the error threshold, the comprehensive performance prediction model and the non-dominated sorting genetic algorithm are used to optimize the performance of the aluminum-based material and the performance of the aluminum-based material, and the composition and process parameters are designed to adjust the composition and process parameters; The adjusted composition and process parameters are issued to the production line through the edge computing system and the 5G mobile network base station for trial production, and the trial production results are fed back to the comprehensive database for online learning and updating of the performance index independent prediction model.

2. The 5G communication and machine learning based intelligent design method of aluminum matrix materials according to claim 1, wherein, Before the real-time acquisition of the acquisition data of the production line, the method further comprises the following steps: For the link on the production line that needs to collect data, the 5G mobile network base station is used to establish a communication connection with the sensors, analyzers and control systems distributed in each work station to realize real-time data collection; Wherein, the link that needs to collect data at least includes smelting, casting, hot rolling, cold rolling, heat treatment, surface treatment and detection.

3. The 5G communication and machine learning based intelligent design method of aluminum matrix materials according to claim 1, wherein, The acquisition data further includes sensor data representing the quality of the aluminum-based material on the production line, and the method further comprises the following steps: Compare the quality index represented by the sensor data with the corresponding data threshold; According to the comparison result, it is determined whether the quality of the aluminum-based material meets the preset quality standard; When the quality of the aluminum-based material does not meet the preset quality standard, an early warning information is sent.

4. The 5G communication and machine learning based intelligent design method of aluminum matrix materials according to claim 1, wherein, The first data set is obtained by the following method: The acquisition data preprocessed by the preprocessing module is used as a data sample, and the preprocessing of the preprocessing module includes data cleaning, denoising, format standardization and feature extraction; Based on the data sample, the composition sample, the process parameter sample and a plurality of performance index samples are obtained to obtain the first data set.

5. The 5G communication and machine learning based intelligent design method of aluminum matrix materials according to any one of claims 1 to 4, characterized in that, The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set.

6. The 5G communication and machine learning based intelligent design method of aluminum matrix materials according to claim 5, wherein, The method comprises the following steps: obtaining a plurality of combinations of the components and the process parameters that can be adjusted; inputting each of the combinations into a plurality of performance index independent prediction models to obtain a plurality of predicted values of the performance indexes of the combination; selecting the combinations whose predicted values of the performance indexes and corresponding target values have errors less than or equal to an error threshold; constructing a comprehensive performance prediction model based on all the selected combinations and predicting the comprehensive performance of each of the combinations by using the comprehensive performance prediction model; obtaining a target set of the combinations whose performance indexes and comprehensive performance are optimized by using a non-dominated sorting genetic algorithm based on the predicted values of the performance indexes of the combinations whose comprehensive performance meets a target performance; and designing the components and the process parameters based on the target set. ​ ​ ​ 7. The 5G communication and machine learning based intelligent design method of aluminum matrix materials according to claim 5, wherein, ​ ​ ​ calculating eigenvalues and corresponding eigenvectors of the correlation coefficient matrix; selecting the eigenvectors corresponding to the m largest eigenvalues from all the eigenvalues and corresponding eigenvectors; obtaining m principal component equations based on the selected eigenvectors corresponding to the m largest eigenvalues, wherein each principal component equation is a linear combination of normalized performance indicators, and the eigenvectors are weights of the normalized performance indicators; calculating a weight ratio of the principal component corresponding to the eigenvalue of the principal component to the sum of the eigenvalues of all the principal components for each principal component; obtaining a comprehensive performance prediction equation based on the sum of the products of the principal component equations and the weight ratios of the principal components; inputting the predicted values of the m performance indicators corresponding to each combination of the n combinations into the comprehensive performance prediction equation to obtain the comprehensive performance of the combination.

8. An aluminum-based material intelligent design device based on 5G communication and machine learning, characterized by, Comprise: a data receiving module configured to receive, based on an edge computing system deployed on an aluminum-based material production line, acquisition data of the production line in real time through a 5G mobile network base station deployed on the production line, and perform local preprocessing on the acquisition data by a preprocessing module of the edge computing system, wherein the acquisition data comprises composition and process parameters; an independent performance prediction module configured to input the received composition and process parameters into a plurality of performance indicator independent prediction models respectively to obtain predicted values of a plurality of performance indicators; wherein the performance indicator independent prediction model is obtained by processing composition samples, process parameter samples and performance indicator samples in a comprehensive database based on machine learning; the comprehensive database is constructed by: a first data set obtained from the production line through the 5G mobile network base station, and a candidate composition obtained based on the calculation results of phase diagram and thermodynamic property calculation based on a thermodynamic database using a phase diagram calculation method, the composition samples are obtained by experimental verification on the candidate composition, the second data set is obtained based on the composition samples, the process parameter samples and a plurality of performance indicator samples, and the comprehensive database is constructed based on the first data set and the second data set; a design adjustment module configured to, in the case that the error between the predicted value of at least one performance indicator and the corresponding target value is greater than an error threshold, optimize a plurality of performance indicators and comprehensive performance of the aluminum-based material through a comprehensive performance prediction model and a non-dominated sorting genetic algorithm, and adjust the composition and the process parameters; a trial feedback module configured to issue the adjusted composition and process parameters to the production line through the edge computing system and the 5G mobile network base station for trial production, and feed back the trial results to the comprehensive database for online learning and updating of the performance indicator independent prediction model.

9. An electronic device, comprising: Comprise: at least one processor; and a memory connected in communication with the at least one processor; wherein The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method for intelligent design of aluminum-based materials based on 5G communication and machine learning according to any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program, when executed by a processor, implements the method for intelligent design of aluminum-based materials based on 5G communication and machine learning according to any one of claims 1 to 7.

Images