Study On Microstructure-property Correlation Of A Fourth-Generation Ni-based Single Crystal Superalloy Based On High-throughput Experiment And Machine Learning

The fourth-generation nickel-based single-crystal(SX)superalloy has become an important alternative material for high-pressure turbine blades of advanced aeroengine with high thrust-to-weight ratio,due to its excellent creep performance and well microstructural stability at high temperature.Fully understanding the relationship between high-temperature creep micro structure and properties under near-service conditions is an important basis for optimizing alloy design and engineering application.In particular,the microstructure evolution and degradation mechanism during creep at high temperature(1000～1150℃)and low stress(<150 MPa)are the focus of research in related fields.However,the conventional experimental method has the problems of high cost and long period,because the fourth-generation SX alloy contains relatively high contents of rare elements Re and Ru,and its creep microstructural evolution at high temperature is complicated and nonlinear,In order to solve the above problems,this study use a fourth-generation nickelbased SX superalloy as the research object,by adopting the concept of Material Genome Engineering,carry out the following researches:(1)According to the microstructural characteristics,a rapid micro structure characterization method for nickel-based SX superalloy during high-temperature creep was developed.(2)Through this method,the microstructural evolution and degradation mechanism of the used superalloy during high temperature creep were studied.(3)Based on the microstructural evolution database,the microstructure-property correlation model of the used superalloy was established by machine learning method.(4)The macroscopic creep model of the used superalloy was established by combining machine learning and θ projection methods.The microstructure-property correlation model was further improved by introducing the creep strains.The rapid microstructure characterization method for investigating the microstructural evolution of nickel-based SX superalloy during high temperature creep developed in this paper integrated the high-throughput experiment,largescale high-resolution characterization and high-throughput quantitative analysis techniques.The high temperature interrupted creep tests were carried out on the variable section specimens with arc surface to acquire the microstructures changed with creep stress continuously.High-resolution SEM with ATLAS module was employed to quickly characterize the large-scale microstructure throughout the universal stress scale.Based on U-Net deep learning algorithm,an automatic dendrite identification model was established to segment the dendrite region quickly and accurately.And then,the γ/γ’ microstructural parameters of dendrite region were continuously quantitated using a logical algorithm.The microstructural evolution during creep at different temperatures and stresses can be effectively studied,which provides a foundation for the study of microstructure-property of nickel-based SX superalloy.The microstructures of the used superalloy under different creep conditions were obtained using the high-throughput experimental method,and the physical microstructural parameters,such as y’ phase volume fraction(Vf),rafting degree(Ω)and raft thickness(D),were quantitatively calculated continuously.On this basis,the microstructural evolution of the used superalloy during high temperature creep was analyzed,and the database of microstructural parameters and creep conditions was established.Moreover,the creep damage mechanism of the used superalloy at high temperature and low stress was analyzed by the characterization of y/y’ two-phase lattice misfit and elemental partitioning behavior and dislocation configuration.Two statistical methods,two-point correlation and principal component analysis(PCA),were used to introduce statistical microstructural parameters(PC1 and PC2)for improving the specificity of creep microstructure features.On this basis,machine learning models of high temperature creep microstructure and properties of the used superalloy were established using the neural network algorithm,to predict the microstructural characteristics under certain creep conditions and the creep conditions corresponding to certain microstructure.The verification experiments showed that the accuracy of the machine learning models was improved effectively by the introduction of statistical microstructural parameters.Based on the database of high temperature creep properties and the physical parameters affecting the high temperature creep properties of nickel-based SX superalloys,the prediction model of high temperature creep properties of nickelbased SX superalloys was established by machine learning method.The creep life of the used superalloy under different creep conditions was predicted by this model.Based on this,the creep curve prediction model of the used superalloy was established by modified θ projection method.Combined with the two macroscopic creep models,the creep strain was introduced to improve the microstructureproperty correlation model of the used superalloy.This work established a series of research methods for the relationship between microstructure and creep property of nickel-based SX superalloy at high temperature,using the concept of Material Genome Engineering and coupling the microstructural evolution and macroscopic creep property,which had been verified by the experimental fourth-generation SX superalloy.The methods are of great significance to effectively reveal the microstructural evolution and establish the relationship between microstructure and property during creep of nickel-based SX superalloys.They also provide theoretical guidance for the subsequent optimization of nickel-based SX superalloys and their service damage assess during industrial application.