Quantitative models of high temperature creep microstructure-property correlation of a nickel-based single crystal superalloy with physical and statistical features

Abstract

The microstructural evolution prediction and operating condition evaluation of nickel-based single crystal (SX) superalloys during creep are of great significance for damage assessment in service. In this paper, quantitative models of crept microstructural evolution of a nickel-based SX superalloy containing Re and Ru were constructed by using the machine learning method with physical and statistical features of microstructure. Firstly, a sequence of high temperature creep tests was conducted on high-throughput specimens with multiple conditions. The physical microstructural features, i.e., volume fraction (Vf), rafting degree (Ω), and rafts thickness (D) of γ′ precipitates of 8 different specimens were quantitated continuously. Secondly, the statistic features were introduced as supplementary to improve the specificity of microstructural features, using the two statistical methods of two-point correlation and principal component analysis (PCA). Then, two machine learning models were constructed through a neural network algorithm, to predict the microstructure under a certain creep condition and evaluate the creep condition with a certain microstructural feature. The validation creep test showed that these two models have good performance. The quantitative models constructed in this study have great significance in the alloy optimization and damage assessment of nickel-based SX superalloys, which can be extended to other SX superalloys.

Keywords

• Nickel-based SX superalloy
• High-throughput creep tests
• Statistical microstructural features
• Microstructure-property correlation
• Artificial neural networks