Data-driven high elastocaloric NiMn-based shape memory alloy optimization with machine learning

Abstract

Shape memory alloys (SMAs) have the potential to improve the efficiency of solid-state refrigeration technology through coupled excitation of multiple thermal effects. Aiming to achieve high elastocaloric NiMn-based SMAs, this paper utilized machine learning to predict the adiabatic temperature change and first-principle calculations to elucidate the mechanism. Based on the optimal XGB Regressor model, the Ni₅₀Mn₃₃Ti₁₇ SMA through directional solidification is predicted to have the highest adiabatic temperature change of 10 K (test temperature = 298 K, applied stress = 300 MPa). In addition, the volume change ratio after martensitic transformation reaches 2.375 % with first-principles calculations, which is expected to provide sufficient entropy and thus obtain an excellent elastocaloric effect. This study provides an available pathway to design and optimize the elastocaloric property of NiMn-based SMAs.