Design of Dual-Phase Steel Based on Active Learning

Dual-phase (DP) steels are an important family of steel grades widely used in the automotive industry, aerospace, ultra-supercritical generating units, etc. Reducing costs throughout the process from raw material preparation to experimental design is a critical challenge that needs to be addressed urgently. This paper develops an effective active machine learning (AL) method to explore and exploit new DP steels with excellent mechanical properties. A simple case of hardness optimization is first reported to validate the reliabilityand efficiency of the AL method. Simultaneous enhancement of strength and plasticityis then realized by fast learning in a vast design space free of Co, finding several desired low-cost DP steels. More importantly, convenient application software has beensuccessfully developed, which has practical significance for the engineering application of the AL method.