Data-driven metallurgical design for high strength low alloy (HSLA) steel

Metallurgical design of microalloyed steel used to be a challenge due to its multivariate nature. Over ten kinds of microalloying elements and multi-step processing routes have complex interactions and different contributions to the final mechanical properties. Data-driven model is able to throw a rapid insight into the composition-processing -property correlation of steel metallurgy in a systematical and efficient way. In this study, a data mining technology, Recursive Partitioning is applied to model the tensile properties of high strength low alloyed (HSLA) steel. The results show that recursive partitioning is able to reveal the complex nonlinear dependence of tensile properties of HSLA steel upon the composition and hot rolling processing parameters. With a relatively simple mathematical structure, Recursive Partitioning can achieve effective performance in predicting the yield strength, ultimate tensile strength, and elongation of steel. In addition, the tree-graph representation of the results provides a powerful multi-dimensional screening tool for searching interesting regions in the composition-processing space, which can be used as a guideline for metallurgical design and further experimental and computational investigation.