Al-Si aluminum alloy high-strength and high-conductivity component design method and component proportion

Abstract

The invention discloses an Al-Si aluminum alloy high-strength and high-conductivity component design method and component proportion, and relates to the technical field of intelligent design of aluminum alloy materials, the Al-Si aluminum alloy high-strength and high-conductivity component design method comprises the steps that S1, a special spoon is used for scooping melt at the temperature of 730 DEG C, after the surface is initially set, the melt is immediately put into cold water for quenching, and a cylindrical test ingot with the diameter of 38 mm and the thickness of 13 mm is prepared; then surface oxide is removed through turning, an electric spark direct-reading spectrometer is adopted for measuring chemical components for three times, an average value is obtained, so that data precision is ensured, 5000 pieces of die casting production data are obtained, and a sample data set containing Al-Si series aluminum alloy multi-element mass fractions and corresponding performance indexes is obtained; the device has the advantages that the development cycle of the high-strength and high-conductivity Al-Si series die-casting aluminum alloy can be shortened, and the lightweight and bearing service requirements of automobile structural parts are met.

Description

Technical Field

[0001] This invention relates to the field of intelligent design technology for aluminum alloy materials, specifically to a high-strength and high-conductivity composition design method and composition ratio for Al-Si aluminum alloys. Background Technology

[0002] Aluminum alloys are widely used in lightweight automotive structural components due to their low density, corrosion resistance, good castability, and excellent thermal and electrical conductivity potential, especially suitable for high-efficiency forming processes such as high-pressure die casting. With the accelerating trend of automotive electrification and integration, structural components face the combined effects of multiple factors such as load, vibration, and temperature rise under more complex service conditions. Localized temperature rises can lead to high-temperature softening, strength reduction, and decreased dimensional stability of the material, thereby affecting structural safety and service reliability. Therefore, aluminum alloys used in automotive structural components not only need to meet lightweight and formability requirements but also need to maintain good thermal and electrical conductivity while ensuring mechanical properties and thermal stability, in order to improve the overall vehicle thermal management efficiency and reduce the risk of localized overheating.

[0003] However, existing high-strength, high-conductivity composition design methods and component ratios for Al-Si aluminum alloys have the following problems in application: The strength and conductivity of metallic materials typically exhibit an inverse relationship. From a microscopic perspective, the conductivity of a material is mainly controlled by the scattering of conduction electrons in the crystal lattice. Structural inhomogeneities such as solute atoms, vacancies, dislocations, grain boundaries, and second-phase particles all enhance electron scattering, leading to increased resistance and decreased conductivity. Correspondingly, commonly used strengthening methods, such as grain refinement, solid solution strengthening, and Orowan strengthening, while improving strength or hardness, often involve increased solute content, dislocation density, or second-phase content, thus adversely affecting conductivity. Therefore, in the alloy composition and microstructure design process for engineering applications, it is necessary to simultaneously consider both conductivity and strength to achieve synergistic optimization. Summary of the Invention

[0004] The purpose of this invention is to provide a high-strength and high-conductivity composition design method and composition ratio for Al-Si aluminum alloys, so as to solve the related problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a high-strength and high-conductivity composition design method and composition ratio for Al-Si aluminum alloys, comprising the following steps: S1: At 730℃, the melt was scooped up with a special spoon and immediately quenched in cold water after the surface had initially solidified to obtain a cylindrical ingot with a diameter of 38mm and a thickness of 13mm. Subsequently, the surface oxide was removed by turning, and the chemical composition was measured three times using an electrical discharge direct reading spectrometer and the average value was taken to ensure data accuracy. 5000 pieces of die casting production data were obtained, including a sample dataset of multi-element mass fractions and corresponding performance indicators of Al-Si aluminum alloys. The performance indicators include at least electrical conductivity (EC) and hardness (HB). S2: Based on thermodynamic software, high-throughput Scheil-Gulliver non-equilibrium solidification calculations are performed on each component point in the training dataset to extract thermodynamic descriptors such as solid solubility, phase volume fraction, solid-liquid interval, and thermal crack sensitivity. These descriptors, together with multi-element mass fractions, constitute the model input features. S3: Using the mass fraction of alloying elements and the thermodynamic characteristics obtained in step S2 as inputs, and EC and HB as outputs, a regression prediction model is established; a PINN multi-task network is used, with a shared backbone network to extract joint characteristics, and the output layers respectively obtain... and Furthermore, a material mechanism consistency constraint term was introduced into the loss function. After the parameters were optimized by grid search, the model was evaluated by cross-validation and combined with indicators such as R² and RMSE. S4: Within the preset component design space, a "component-performance" surrogate model is constructed using Bayesian optimization, with conductivity and hardness as multi-objective optimization targets. The next round of candidate component points is selected through a data acquisition function, and the surrogate model is iteratively updated to generate a candidate solution set and extract the Pareto front, wherein the data acquisition function is used. S5: Select the optimal compromise on the Pareto front as the recommended candidate composition scheme, and conduct a feasibility analysis on the solidification and phase transformation behavior of the designed alloy to determine whether it meets the casting process requirements. S6: Perform thermodynamic calculations on the recommended candidate component schemes to obtain relevant thermodynamic descriptors, which serve as the basis for mechanism explanation and process feasibility judgment. Subsequently, conduct trial production and performance testing on the candidate schemes, and backfill the measured results into the sample dataset to update the positive prediction model and drive the Bayesian optimization process for iterative optimization.

[0006] As a preferred embodiment of the Al-Si aluminum alloy high-strength and high-conductivity composition design method and composition ratio of the present invention, abnormal samples are removed in step S1 using the interquartile range rule, where IQR = Q3 - Q1. When the performance index value of a sample is less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR, the sample is judged as abnormal and removed, where Q1 and Q3 are the first and third quartiles of the corresponding performance index, respectively.

[0007] As a preferred embodiment of the Al-Si aluminum alloy high strength and high conductivity composition design method and composition ratio of the present invention, the Bayesian optimization in step S4 adopts Gaussian process regression as a surrogate model, the acquisition function is used to evaluate candidate composition points, and the candidate composition points that make the acquisition function take the maximum value are selected to enter the next round of iteration; wherein the acquisition function is preferably a multi-objective expectation improvement (MOEI).

[0008] As a preferred embodiment of the Al-Si aluminum alloy high strength and high conductivity composition design method and composition ratio of the present invention, the robust performance index is q_max, where q_max is the upper bound predicted value of the performance output by the regression prediction model under a preset confidence level. In step S4, HB_qmax and EC_qmax are used as the objective functions of multi-objective optimization, and Pareto solution set is generated under the condition of satisfying the preset engineering threshold constraint.

[0009] As a preferred embodiment of the Al-Si aluminum alloy high strength and high conductivity composition design method and composition ratio of the present invention, in step S5, the optimal compromise point is selected as the recommended candidate composition scheme on the Pareto front. The selection of the optimal compromise point is based on the position where the curvature of the Pareto curve reaches its maximum after normalization of the objective function, or the position where the distance from the candidate point to the ideal point reaches its minimum.

[0010] As a preferred embodiment of the Al-Si system aluminum alloy high strength and high conductivity composition design method and composition ratio of the present invention, the thermodynamic and solidification process calculation in step S6 is performed using Pandat software, which includes phase equilibrium calculation and Scheil non-equilibrium solidification calculation, and outputs hot cracking tendency factor, phase volume fraction and solid solubility results for screening candidate composition schemes.

[0011] As a preferred embodiment of the Al-Si aluminum alloy high-strength and high-conductivity composition design method and composition ratio of the present invention, Si is 6–10 wt.%, Mg is 0–0.8 wt.%, Cu is 0–0.95 wt.%, Fe is 0–1.5 wt.%, Mn is 0–0.3 wt.%, Zn is 0–0.2 wt.%, Sr is 0–0.10 wt.%, and the remainder is Al and unavoidable impurities.

[0012] As a preferred embodiment of the Al-Si aluminum alloy high-strength and high-conductivity composition design method and composition ratio of the present invention, the aluminum alloy satisfies the preset casting process feasibility constraints, the constraints including at least: the solid-liquid interval ΔT obtained by Scheil-Gulliver non-equilibrium solidification calculation does not exceed the solid-liquid interval threshold ΔT_th, and the hot cracking tendency factor does not exceed the hot cracking sensitivity threshold HTI_th; wherein ΔT_th and HTI_th are respectively taken from the third quartile Q3 (or preset quartile) of the corresponding index distribution in the training dataset.

[0013] The present invention discloses a high-strength and high-conductivity composition design method for Al-Si aluminum alloys and the beneficial effects of the composition ratio: 1. This invention can rapidly screen candidate composition points that meet the synergistic goals of electrical conductivity and hardness within a preset alloy composition range, accelerating the design and development of novel high-strength, high-conductivity die-cast aluminum alloys. Specifically, this invention acquires 5000 pieces of die-casting production data, including the mass fraction of each alloy element and target performance data such as electrical conductivity and hardness. Based on computational thermodynamics software, Scheil non-equilibrium solidification simulation is performed on the sample composition to obtain thermodynamic and solidification process parameters such as solid solubility, solid-liquid temperature difference, phase volume fraction, and hot cracking tendency factor, forming a complete dataset. On this basis, box plots and interquartile range (IQR) rules are used to remove outlier samples, and manual verification is performed to complete data cleaning and standardization. Using elemental composition and thermodynamics as input features, and electrical conductivity and hardness as output targets, a PINN prediction model is trained, and grid search and cross-validation are used for model optimization. Furthermore, within the preset composition design space, the regression prediction model is used as a performance evaluator. A surrogate model is constructed using Gaussian process regression, and the improved acquisition function MOEI is combined with multi-objective expectation to iteratively select points, generate candidate solution sets, and extract the Pareto front, thereby obtaining a compromise optimal candidate composition scheme that balances conductivity and hardness. The preset alloy composition content range is: Si mass percentage of 6-10 wt.%, Mg mass percentage of 0-0.8 wt.%, Cu mass percentage of 0-0.95 wt.%, with Al as the balance. The contents of other elements are fixed to preset values ​​or limited to the statistical range of the dataset. Since this invention introduces thermodynamics simultaneously during the reverse search process, it can effectively reduce the output of candidate schemes that are metallurgically infeasible or lack manufacturability due to relying solely on data-driven approaches, and improve the engineering feasibility and interpretability of the screening results. Therefore, this invention can shorten the development cycle of high-strength, high-conductivity Al-Si die-cast aluminum alloys, reduce trial-and-error costs, and improve the probability of comprehensive performance compliance of candidate alloy schemes, meeting the requirements of lightweight and load-bearing service for automotive structural components.

[0014] 2. In this invention, when reselecting the optimal compromise composition point, the measured conductivity and hardness data of each of the proposed prototype schemes are first added to the initial dataset as new data points. Then, in conjunction with step S3, the regression prediction model is updated and trained. Finally, step S4 is used to re-execute Bayesian optimization iteration within the preset alloy composition content range and extract the Pareto front, thereby redetermining the optimal compromise point. Through this dynamic adjustment and optimization, the preferred composition scheme that balances conductivity and hardness within the preset alloy composition content range is ensured to be selected, so that the final Al-Si die-cast aluminum alloy meets the high requirements of practical applications. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall process of a multi-objective reverse design method for high-strength and high-conductivity die-cast aluminum alloys according to the present invention; Figure 2 This is the model performance evaluation result of the PINN ablation comparison in Example 2, showing the consistency between the predicted and measured values ​​of hardness HB and conductivity EC and the ±RMSE error band. Figure 3 This is a box plot of the relative error distribution of the PINN ablation comparison in Example 2, used to compare the statistical distribution of the prediction errors of HB and EC under different constraint settings; Figure 4 This is the physical consistency test result in Example 2, showing the box plot of the residual distribution between the predicted value and the physical model; Figure 5 This is the result diagram of the solidification and phase transformation feasibility analysis of the recommended candidate components in Example 2, showing the evolution curves of phase fraction, elemental distribution and related thermodynamic quantities as a function of temperature under Scheil non-equilibrium solidification conditions. Figure 6 This is the response surface and two-dimensional projection distribution map of the thermodynamic characteristic indexes in the design space in this embodiment 2, which is used to evaluate the impact of compositional fluctuations on solidification behavior and process window and to assist in screening stable regions. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1, such as Figure 1 As shown, the present invention provides a technical solution: a method for designing and proportioning high-strength and high-conductivity Al-Si aluminum alloys, comprising the following steps: S1: At 730℃, the melt was scooped up with a special spoon and immediately quenched in cold water after the surface had initially solidified to obtain a cylindrical ingot with a diameter of 38mm and a thickness of 13mm. Subsequently, the surface oxide was removed by turning, and the chemical composition was measured three times using an electrical discharge direct reading spectrometer and the average value was taken to ensure data accuracy. 5000 pieces of die casting production data were obtained, including a sample dataset of multi-element mass fractions and corresponding performance indicators of Al-Si aluminum alloys. The performance indicators include at least electrical conductivity (EC) and hardness (HB). In step S1, the interquartile range rule is used to remove outlier samples, where IQR = Q3 - Q1. When the performance index value of a sample is less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR, the sample is judged as outlier and removed. Q1 and Q3 are the first and third quartiles of the corresponding performance index, respectively. Step S1 involves acquiring 5000 pieces of die-casting production data. This sample data includes at least the mass fraction of each alloying element and its corresponding electrical conductivity and hardness. All samples were sourced from the machine-side furnace. The sampling and sample preparation process was as follows: The melt was scooped using a special spoon at 730℃, and immediately quenched in cold water after initial surface solidification to obtain cylindrical ingots with a diameter of 38mm and a thickness of 13mm. Subsequently, the surface oxide was removed by machining, and the chemical composition was measured three times using an electrical discharge spectrometer, with the average value taken to ensure data accuracy. The entire process strictly followed the company's standard operating procedures, and the influence of differences in solidification conditions on composition, hardness, and electrical conductivity can be approximately ignored. A portion of the sample set is shown in Table 1.

[0018] Furthermore, the mass percentage of Si is 6.65 wt.% to 13.19 wt.%, the mass percentage of Fe is 0.637 wt.% to 0.996 wt.%, the mass percentage of Cu is 0.0017 wt.% to 0.837 wt.%, and the mass percentage of Mn is 0.0052 wt.% to 0.487 wt.%. Based on the computational thermodynamic software Pandat, a high-throughput Scheil-Gulliver nonequilibrium solidification simulation was performed on the sample composition. The initial temperature was set above the liquidus temperature and then cooled to room temperature. The output included at least: (1) the solid solution content of each alloying element in the matrix phase, (2) the solid-liquid temperature difference, (3) the volume fraction of the key phase, and (4) the indicators related to hot cracking sensitivity. Table 1 Dataset of Cast Aluminum Alloys S2: Based on thermodynamic software, high-throughput Scheil-Gulliver non-equilibrium solidification calculations are performed on each component point in the training dataset to extract thermodynamic descriptors such as solid solubility, phase volume fraction, solid-liquid interval, and thermal crack sensitivity. These descriptors, together with multi-element mass fractions, constitute the model input features. S2: Based on thermodynamic software, high-throughput Scheil-Gulliver non-equilibrium solidification calculations are performed on each component point in the training dataset to extract thermodynamic descriptors such as solid solubility, phase volume fraction, solid-liquid interval, and thermal crack sensitivity. These descriptors, together with multi-element mass fractions, constitute the model input features. Step S2: The dataset obtained in Step S1 is deduplicated, and outlier samples are identified and removed using box plots and the interquartile range (IQR) rule. Sample values ​​exceeding Q1 - 1.5 × IQR or Q3 + 1.5 × IQR are identified as outliers and removed. These outliers are manually reviewed to reduce noise impact caused by measurement fluctuations, data entry errors, or process anomalies. Subsequently, the input features are standardized, preferably using Z-score standardization to reduce the impact of different units on model training. Optionally, the dataset can be divided into training and testing sets, or 5-fold cross-validation can be used to evaluate the model's generalization performance.

[0019] Example 2, as Figure 1-6 As shown, this invention provides a technical solution: a reverse design method for high-strength and high-conductivity Al-Si die-cast aluminum alloys based on deep learning and Bayesian optimization, and incorporating thermodynamic parameters. The method includes S3: using the mass fraction of alloying elements and the thermodynamic characteristics obtained in step S2 as inputs, and EC and HB as outputs, a regression prediction model is established; a PINN multi-task network is used, sharing the backbone network to extract joint characterizations, and the output layers respectively obtain... and Furthermore, a material mechanism consistency constraint term was introduced into the loss function. After the parameters were optimized by grid search, the model was evaluated by cross-validation and combined with indicators such as R² and RMSE.

[0020] Step S3: In this embodiment, a PINN-style multi-task network is preferably used as the regression prediction model: a shared backbone network is used to extract joint characteristics, and the output layers provide hardness and conductivity respectively. Material mechanism constraints are introduced into the loss function to ensure that the prediction results are consistent with the superposition laws of solid solution strengthening, second phase contribution, and resistivity. in, For the first Solid solubility of each element in the matrix. For the first Phase fraction of each phase, It is a non-linear exponent used to describe a diminishing marginal trend. These are coefficients to be learned or calibrated. Solute element The resistivity contribution coefficient. For the second phase The resistivity contribution coefficient. (From...) Figure 2 As shown, after adding the physical consistency term, the scatter points are closer to the main diagonal. The scheme shows compact point clouds and smaller error band dispersion on both HB and EC tasks, indicating that the dual physical constraints have complementary benefits for the dual objectives. After the model training is completed, the model is comprehensively evaluated by indicators such as the coefficient of determination R² and root mean square error RMSE through 5-fold cross-validation. It can be seen that the R² of HB increased from 0.806 to 0.832 and the R² of HB increased from 0.764 to 0.942, which means that the model training is complete. S4: Within the preset component design space, a "component-performance" surrogate model is constructed using Bayesian optimization, with conductivity and hardness as multi-objective optimization objectives. The next round of candidate component points is selected through a data acquisition function, and the surrogate model is iteratively updated to generate a candidate solution set and extract the Pareto front. The data acquisition function is preferably the Multi-Objective Expected Improvement (MOEI). Step S4: Within the preset component design space, using the regression prediction model from Step S3 as a performance evaluator, Bayesian optimization is employed to perform a multi-objective inverse search. The Bayesian optimization uses a Gaussian surrogate model to construct a "component-performance" surrogate function, and selects candidate component points for the next round through a collection function, preferably MOEI, to balance exploration and development, iteratively generating a set of candidate components, and forming a candidate solution set and extracting the Pareto front under engineering threshold constraints. Detailed model parameters are shown in Table 2.

[0021] Table 2 Kernel function and hyperparameters of the regression model The composition design space is defined as follows: Si 6–10 wt.%, Mg 0–0.8 wt.%, Cu 0–0.95 wt.%, Fe 0–1.5 wt.%, Mn 0–0.3 wt.%, Zn 0–0.2 wt.%, Sr 0–0.1 wt.%, and Al as the balance. Other trace elements are fixed at preset values ​​or limited to the statistical range of the dataset to ensure the industrial manufacturability of the candidate solution.

[0022] Ultimately, the present invention designed the optimal alloy formula within the stated scheme: Al-7.49%Si-0.26%Mg-1.36%Fe-0.19%Mn-0.17%Zn-0.03%Cu-0.08%Sr; S5: Select the optimal compromise on the Pareto front as the recommended candidate composition scheme, and conduct a feasibility analysis on the solidification and phase transformation behavior of the designed alloy to determine whether it meets the casting process requirements. Step S5: Select the optimal compromise on the Pareto front as the recommended candidate composition scheme, and conduct a feasibility analysis of the solidification and phase transformation behavior of the designed alloy to determine whether it meets the casting process requirements. For example... Figure 5 As shown in Table 3, the candidate composition initially exhibits a liquid phase coexisting with the Fe-rich intermetallic compound Al13Fe4 in the high-temperature range. Subsequently, as the temperature decreases, Al15(FeMn)3Si2, Al2Si2Sr, and β-AlFeSi are gradually formed. When the temperature continues to decrease to approximately 573.3–609.8℃, Fcc begins to appear, indicating that the matrix phase has entered the main solidification stage. In the approximately 556.0–573.1℃ range, Si further appears and coexists with Fcc, exhibiting typical eutectic solidification characteristics of the Al-Si system. Based on the above solidification path and phase composition results, the casting suitability of the candidate alloy can be determined, such as... Figure 4 As shown, if the phase content during the solidification interval is within a controllable range, and there is no phase transformation behavior leading to severe hot cracking at the end of solidification, then the candidate composition scheme is considered to meet the process feasibility requirements for die casting. Conversely, the precipitation trend of the second phase and its temperature range can be used as constraints, and the results can be fed back to step S4 for screening and iterative optimization of the candidate solution set.

[0023] The solidification path of the alloy formulation S6: Perform thermodynamic calculations on the recommended candidate component schemes to obtain relevant thermodynamic descriptors, which serve as the basis for mechanism explanation and process feasibility judgment. Subsequently, conduct trial production and performance testing on the candidate schemes, and backfill the measured results into the sample dataset to update the positive prediction model and drive the Bayesian optimization process for iterative optimization.

[0024] In step S4, the Bayesian optimization uses Gaussian process regression as a surrogate model. The acquisition function is used to evaluate candidate component points, and the candidate component points that maximize the acquisition function are selected to enter the next iteration. The acquisition function is preferably Multi-Objective Expectation Improvement (MOEI), and the robust performance index is q_max, which is the upper bound prediction value of the performance output by the regression prediction model under a preset confidence level. In step S4, HB_qmax and EC_qmax are used as the objective functions of multi-objective optimization, and Pareto solution sets are generated under the condition of satisfying the preset engineering threshold constraints.

[0025] In step S5, the optimal compromise point is selected as the recommended candidate component scheme on the Pareto front. The selection of the optimal compromise point is based on the position where the curvature of the Pareto curve reaches its maximum after normalization of the objective function, or the position where the distance from the candidate point to the ideal point reaches its minimum.

[0026] The thermodynamic and solidification process calculations in step S6 were performed using Pandat software, which included phase equilibrium calculations and Scheil non-equilibrium solidification calculations. The results of thermal cracking tendency factor, phase volume fraction and solid solubility were output for screening candidate composition schemes.

[0027] The alloy comprises 6–10 wt.%, 0–0.8 wt.%, 0–0.95 wt.%, 0–1.5 wt.%, 0–0.3 wt.%, 0–0.2 wt.%, and 0–0.10 wt.%, with the remainder being Al and unavoidable impurities. The aluminum alloy meets the preset feasibility constraints of the casting process. The constraints include at least the following: the solid-liquid interval ΔT obtained by Scheil-Gulliver nonequilibrium solidification calculation does not exceed the solid-liquid interval threshold ΔT_th, and the hot cracking tendency factor does not exceed the hot cracking sensitivity threshold HTI_th; wherein ΔT_th and HTI_th are respectively taken from the third quartile Q3 (or the preset quartile) of the corresponding index distribution in the training dataset.

[0028] The initial experimental performance data includes measured data on the electrical conductivity and hardness of the alloy with the preset alloy composition. Specifically, the alloy raw materials used include pure aluminum, pure Mg, and pure Zn (all with a purity of 99.85%), as well as various intermediate alloys such as Al-20Si, Al-10Fe, Al-10Mn, and Al-40Cu. All metals are first weighed according to the designed mass fraction, and the surface oxide scale is removed by light grinding with a grinding wheel. After being immersed in alcohol, they are dried to ensure cleanliness. A clay crucible is selected as the melting vessel: the crucible is preheated to 300°C in a box-type resistance furnace and held at that temperature for 30 minutes. Then, a water-based boron nitride coating is uniformly coated on the inner wall and baked further to ensure that the coating is dry and dense. Afterward, pure aluminum is placed in the crucible and heated to about 700°C in the furnace until it is completely melted. Then, Al-20Si is added, and the mixture is stirred in a circular direction for 3 minutes using a dry graphite rod to promote dissolution and uniform composition. When the temperature reached approximately 720°C, Al–10Fe, Al–10Mn, and Al–40Cu were added in batches, with continuous stirring for 3 minutes after each addition to ensure full diffusion of the alloying elements. After the master alloy was completely melted, the slag on the surface was gently skimmed off with a graphite rod and the mixture was stirred again to further homogenize the melt composition. The melt was then cooled to approximately 730°C for argon refining and degassing: high-purity argon (99.99%) was delivered to an alumina lance via a hose, with the lance inserted obliquely 10–15 cm below the liquid surface. The flow rate was adjusted to allow bubbles to escape evenly and carry impurities to the surface. The degassing process lasted 30 minutes. After degassing, the melt was allowed to stand for 5 minutes to further separate the slag, and then thoroughly skimmed off the slag with casting tongs to ensure the purity of the melt. Finally, to ensure consistency in the alloy solidification conditions, the mold, melting temperature, and casting temperature used in the experiment were all consistent with the sampling methods used in the dataset. For post-forming performance testing and microscopic characterization analysis, the obtained alloy samples were first sliced, and the surface of the samples was successively polished using metallographic sandpaper of 240 mesh, 400 mesh, 600 mesh, 800 mesh, 1500 mesh, and 2000 mesh, and then cleaned and dried. Electrical conductivity was measured using an eddy current conductivity meter (Helmut Fischer SIGMASCOPE SMP10). Hardness was measured using a Buehler Wilson BH3000 Brinell hardness tester. Composition was analyzed using inductively coupled plasma mass spectrometry (ICP-MS) (Agilent 7900).

[0029] Table 3 Comparison of Predicted and Actual Measured Values Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A method for designing high-strength and high-conductivity compositions in Al-Si aluminum alloys, characterized in that: Includes the following steps: S1: At 730℃, the melt was scooped up with a special spoon and immediately quenched in cold water after the surface had initially solidified to obtain a cylindrical ingot with a diameter of 38mm and a thickness of 13mm. Subsequently, the surface oxide was removed by turning, and the chemical composition was measured three times using an electrical discharge direct reading spectrometer and the average value was taken to ensure data accuracy. 5000 pieces of die casting production data were obtained, including a sample dataset of multi-element mass fractions and corresponding performance indicators of Al-Si aluminum alloys. The performance indicators include at least electrical conductivity EC and hardness HB. S2: Based on thermodynamic software, high-throughput Scheil-Gulliver non-equilibrium solidification calculations are performed on each component point in the training dataset to extract thermodynamic descriptors such as solid solubility, phase volume fraction, solid-liquid interval, and thermal crack sensitivity. These descriptors, together with multi-element mass fractions, constitute the model input features. S3: Using the mass fraction of alloying elements and the thermodynamic characteristics obtained in step S2 as inputs, and EC and HB as outputs, a regression prediction model is established; a PINN multi-task network is used, with a shared backbone network to extract joint characteristics, and the output layers respectively obtain... and Furthermore, a material mechanism consistency constraint term was introduced into the loss function. After the parameters were optimized by grid search, the model was evaluated by cross-validation and combined with indicators such as R² and RMSE. S4: Within the preset component design space, a "component-performance" surrogate model is constructed using Bayesian optimization, with conductivity and hardness as multi-objective optimization targets. The next round of candidate component points is selected through a collection function, and the surrogate model is iteratively updated to generate a candidate solution set and extract the Pareto front, wherein the collection function is used. S5: Select the optimal compromise on the Pareto front as the recommended candidate composition scheme, and conduct a feasibility analysis on the solidification and phase transformation behavior of the designed alloy to determine whether it meets the casting process requirements. S6: Perform thermodynamic calculations on the recommended candidate component schemes to obtain relevant thermodynamic descriptors, which serve as the basis for mechanism explanation and process feasibility judgment. Subsequently, conduct trial production and performance testing on the candidate schemes, and backfill the measured results into the sample dataset to update the positive prediction model and drive the Bayesian optimization process for iterative optimization.

2. The method for designing high-strength and high-conductivity Al-Si aluminum alloys according to claim 1, characterized in that: In step S1, the interquartile range rule is used to remove outlier samples, where IQR = Q3 - Q1. When the performance index value of a sample is less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR, the sample is judged as outlier and removed. Q1 and Q3 are the first and third quartiles of the corresponding performance index, respectively.

3. The method for designing high-strength and high-conductivity compositions of Al-Si aluminum alloys according to claim 1, characterized in that: In step S4, the Bayesian optimization uses Gaussian process regression as a surrogate model, the acquisition function is used to evaluate candidate component points, and the candidate component points that make the acquisition function take the maximum value are selected to enter the next round of iteration.

4. The method for designing high-strength and high-conductivity Al-Si aluminum alloys according to claim 1, characterized in that: The performance index is q_max, which is the upper bound predicted value of the performance output by the regression prediction model under a preset confidence level. In step S4, HB_qmax and EC_qmax are used as the objective functions of multi-objective optimization, and Pareto solution set is generated under the condition of satisfying the preset engineering threshold constraint.

5. The method for designing high-strength and high-conductivity compositions of Al-Si aluminum alloys according to claim 1, characterized in that: In step S5, the optimal compromise point is selected as the recommended candidate component scheme on the Pareto front. The selection of the optimal compromise point is based on the position where the curvature of the Pareto curve reaches its maximum after normalization of the objective function, or the position where the distance from the candidate point to the ideal point reaches its minimum.

6. The method for designing high-strength and high-conductivity compositions of Al-Si aluminum alloys according to claim 1, characterized in that: The thermodynamic and solidification process calculations in step S6 were performed using Pandat software, which included phase equilibrium calculations and Scheil non-equilibrium solidification calculations. The results of thermal cracking tendency factor, phase volume fraction and solid solubility were output for screening candidate composition schemes.

7. A method for designing high-strength and high-conductivity Al-Si aluminum alloy compositions according to any one of claims 1-6, characterized in that: Its composition ratio is: Si 6–10 wt.%, Mg 0–0.8 wt.%, Cu 0–0.95 wt.%, Fe 0–1.5 wt.%, Mn 0–0.3 wt.%, Zn 0–0.2 wt.%, Sr 0–0.10 wt.%, with the remainder being Al and unavoidable impurities.

8. The method for designing high-strength and high-conductivity compositions of Al-Si aluminum alloys according to claim 7, characterized in that: Its composition ratio: The aluminum alloy meets the preset casting process feasibility constraints, the constraints include at least the following: the solid-liquid interval ΔT obtained by Scheil-Gulliver non-equilibrium solidification calculation does not exceed the solid-liquid interval threshold ΔT_th, and the hot cracking tendency factor does not exceed the hot cracking sensitivity threshold HTI_th; wherein ΔT_th and HTI_th are respectively taken from the third quartile Q3 of the corresponding index distribution in the training dataset.

Images