Machine learning interphase precipitation behavior of Ti micro-alloyed steel guided by physical metallurgy principle

Abstract

Nano-sized interphase precipitates, which form in ordered rows, are critical for high strength low alloy (HSLA) steels, in achieving the desired strength needed for downgauging for light-weighting automotive structures. The occurrence of interphase precipitation depends on many interrelated parameters, such as alloy chemical composition, temperature, processing parameters and crystallography. In this paper, we use data analysis based on machine learning algorithm: decision tree to predict whether interphase precipitation can occur. Due to the high strength of interphase precipitation and the minimum alloy content is one of the important goals pursued of iron and steel enterprises. Therefore, under the condition that interphase precipitation occurs, the chemical composition of the alloy can be reduced by using interphase precipitation. Guided by the physical metallurgy principle, this paper transforms the processing parameters into physical metallurgy parameters such as grain size, stored deformation energy, ferrite phase transformation temperature, and decision tree (DT) algorithm was used to model and verify whether interphase precipitation can occur. Under the condition of interphase precipitation, the support vector machine (SVM) models of interphase precipitation characteristic values were established. Based on the DT model and the SVM model, the particle swarm optimization (PSO) algorithm was used for the reduction design of the alloy, and the chemical composition can be reduced according to the target strengthening strength. The results of alloy reduction have been verified by experiments, which shows that the alloy can be reduced by using interphase precipitation.

Keywords

• Micro-alloyed steel
• Interphase precipitation
• Characteristic value
• Machine learning