Study On The Novel Al-Li Alloys With High Specific Modulus Based On Machine Learning Method

With the rapid development of lightweighting in high-end manufacturing,the problem of insufficient structural stiffness of lightweight and thin-walled components has become more apparent,and the elastic modulus has become an important indicator for measuring lightweight structural materials.Al-Li alloy(Aluminum-Lithium Alloy)has high strength,high stiffness and low density,and is recognized as an ideal structural material for aerospace.High specific modulus Al-Li alloys have become the focus of research and hot topics in the field of aluminum alloys.In recent years,machine learning,as a data-driven technology,has become an important means of improving the efficiency of developing new materials and has received widespread attention.Based on database and machine learning technology,this study conducts research on the composition design of a new high specific modulus Al-Li alloy,which can provide technical reference for the development of high-performance AlLi alloys for aerospace.In this paper,a machine learning method is used to carry out a datadriven design of the composition and properties of alloy materials based on the design concept of Al-Li alloy composition,and a high-precision Al-Li alloy specific modulus dataset is established.A highly efficient integrated algorithm for predicting the specific modulus of Al-Li alloys is trained,and combined with experimental verification,a new alloy composition design is carried out for optimizing the specific modulus of Al-Li alloys and element substitution.The main research content and achievements include:In this study,the design and preparation of a new high-specific modulus Al-Li-Cu-Mg alloy was realized through the establishment of the specific modulus dataset of aluminumlithium alloy and the research of the integrated algorithm of the specific modulus of highefficiency aluminum alloy.On the basis of collecting 145 elastic modulus datasets of aluminum-lithium alloys,the features were constructed based on the content of alloying elements,and the model was trained by AdaBoost Regressor,Gradient Boosting Regressor,Random Forest and other integrated algorithms,and the coefficient of determination(R2)and Mean Squared Error(MSE)were used as evaluation methods,and the gradient enhancement tree was found to have the highest prediction accuracy through 10-fold crossvalidation.After optimizing the model parameters,the evaluation scores for the training set and the test set are R2 of 0.90 and 0.81,respectively.Six high-specific modulus Al-Li-Cu-Mg alloys were predicted by using the gradient enhancement tree model of parameter optimization,and the measured results were found to be close to the predicted results after experimental validation.The maximum specific modulus of the design alloy reached 31.69 GPa/(g·cm-3),which was about 4%higher than the maximum specific modulus of the comparison data set.Domain knowledge and expert experience guide to achieve the comprehensive optimization of specific strength and specific modulus,predict that Al-2.32 Li-1.44 Cu-2.78 Mg-0.3 Ag-0.3 Mn-0.1 Zr alloy has the best specific strength and specific modulus matching,compared with 2195-T8 alloy,the specific modulus is increased by 12.6%under the condition that the specific strength is basically the same.In this study,the composition optimization of Al-2.32 Li-1.44 Cu-2.78 Mg-0.3 Ag-0.3 Mn-0.1 Zr alloy was extended by setting the solid solubility threshold of the elements in the aluminum alloy to select the alloying element that is expected to improve the specific modulus of the alloy.A specific modulus prediction model of Al-Li alloy with stronger generalization ability was constructed,and an alloy composition optimization strategy was formulated based on the composition of the benchmark alloy,and the specific modulus of the virtual alloy search space containing alternative elements was predicted.Among the predicted alloys,alloy with a substitution of 3.40 wt.%of Ho elements have the optimal modulus of elasticity and specific modulus,which is higher than the elastic modulus of the benchmark alloy,and the maximum specific modulus of the design alloy reaches 31.62 GPa/(g·cm-3),which is comparable to the specific modulus level of the benchmark alloy.The alloy with a substitution amount of Ga element of 0.96 wt.%has the best microstructure treatment effect,and its optimal tensile strength(523 MPa)and specific modulus[31.53 GPa/(g·cm-3)]are comparable to the benchmark alloy.