Prediction Of Mechanical Properties Of Fe-Mn-Al-C Low-density High-strength Stee

Fe-Mn-Al-C low-density high-strength steel,with its excellent mechanical properties,good impact resistance and low density,will become the main material for future automotive steels and is one of the main research directions for implementing the goal of automotive lightweighting.However,the addition of large amounts of Al(up to 13%)and Mn(up to 32%)makes it significantly different from the development and design of steels for general use in terms of smelting,forging,microstructure and plastic deformation mechanism,and the use of traditional trial-and-error method experiments has greatly slowed down its development progress.Machine learning techniques have now been widely applied to predict material properties and guide the design and development of new materials,greatly reducing the time and cost of material development.Therefore,in this paper,machine learning techniques are used to predict the mechanical properties of Fe-Mn-Al-C low-density high-strength steels from the perspectives of both compositional processes and atomic features,and the effects of relevant features on mechanical properties are further investigated,with the main research work as follows.(1)To address the problem that there is no publicly available dataset for Fe-MnAl-C low-density high-strength steels that can be used for machine learning prediction,we collect composition-process-property data of low-density high-strength steels based on published research literature on this system and construct a dataset that can be used for machine learning.Based on the collected original dataset,atomic-scale features of the relevant elements are calculated to construct a relevant atomic-scale feature dataset.The constructed dataset provides data support for the prediction of mechanical properties of Fe-Mn-Al-C low-density high-strength steels based on machine learning.(2)Aiming at the problem that conventional experiments cannot achieve direct prediction of mechanical properties by composition process,a machine learning model for predicting the mechanical properties of Fe-Mn-Al-C low-density high-strength steel by composition process was designed.The machine learning model uses a genetic algorithm to optimize the BP neural network,and after verification of experimental data,it is found that the constructed prediction model can accurately predict the mechanical properties of its steel of this system.Further,the prediction model is used to analyze the effects of changes in the content of Mn,Al and C elements and changes in solid solution temperature on the mechanical properties of low-density highstrength steel.(3)In order to reveal the relationship between the microstructure and macroscopic properties of Fe-Mn-Al-C low-density high-strength steels,the effects of atomic features on two mechanical properties(yield strength and total elongation)of lowdensity high-strength steels were investigated,and a machine learning model for atomic features to predict their mechanical properties was developed.First,the computed atomic features are filtered by Pearson correlation and then recursively eliminated.Then,the obtained strongly correlated features are constructed to support the prediction model of vector regression,and the mechanical properties of Fe-Mn-AlC low-density high-strength steels are predicted using the strongly correlated features,which provides a feasible method to reveal the relationship between microstructure and macroscopic properties of steel materials.This paper focuses on the problem that the Fe-Mn-Al-C low-density high-strength steel is difficult to efficiently investigate the composition-process-property relationship by the traditional experimental trial-and-error method,and proposes the use of machine learning methods to investigate the composition-process-property relationship of this system of steel,so as to accelerate the research progress of Fe-MnAl-C low-density high-strength steel.