Accelerated learning and co-optimization of elastocaloric effect and stress hysteresis of elastocaloric alloys  

作　　者：Shi-Yu He Fei Xiao Rui-Hang Hou Shun-Gui Zuo Ying Zhou Xiao-Rong Cai Zhu Li Yan-Ming Wang Aysu Catal-Isik Enrique Galindo-Nava Xue-Jun Jin 

机构地区：[1]State Key Lab of Metal Matrix Composite,School of Materials Science and Engineering,Shanghai Jiao Tong University,Shanghai,200240,China [2]Materials Genome Initiative Center,Shanghai Jiao Tong University,Shanghai,200240,China [3]Institute of Medical Robotics,Shanghai Jiao Tong University,Shanghai,200240,China [4]Department of Materials Science and Engineering,Graduate School of Engineering,Osaka University,Osaka,565-0871,Japan [5]University of Michigan-Shanghai Jiao Tong University Joint Institute,Shanghai Jiao Tong University,Shanghai,200240,China [6]Department of Mechanical Engineering,University College London,Torrington Place,London,WC1E 7JE,UK

出　　处：《Rare Metals》2024年第12期6606-6624,共19页稀有金属(英文版)

基　　金：financially supported by the National Natural Science Foundation of China(Nos.52031005 and 52211530096);Shanghai Academy of Spaceflight Technology Joint Research Fund(No.USCAST2023-19);the Science and Technology Commission of Shanghai Municipality(No.20DZ2220400);the Open Project of the State Key Laboratory of Robotics(No.2023-019);the Equipment Development Department Huiyan Action;the Osaka University's International Joint Research Promotion Program;the Royal Society for the funding,via the International Exchanges program(No.IECNSFC211187)。

摘　　要：The development of high-performance advanced elastocaloric alloys is crucial for the implementation of elastocaloric refrigeration.Here,we present a machine learning(ML)framework to accelerate the development of novel elastocaloric alloys with large adiabatic temperature change(ΔT_(ad))and low stress hysteresis(Δσ_(hy)).The comprehensive framework comprises database construction,feature selection,model construction,alloy design and validation,and model interpretation.Features are selected according to the physical attributes they represent.Properties that may reflect the compatibility between parent and product phases,lattice distortion and the free energy in the alloy are considered in the model.Among them,the key features are selected by recursive feature elimination and exhaustive search methods.The trained models in combination with the Bayesian optimization method are exploited to achieve multi-objective optimization.According to the results,a newly designed elastocaloric alloy shows a large adiabatic temperature change of 15.2 K and low average stress hysteresis of70.3 MPa at room temperature,which is consistent with our predictions.The predictions of our ML model are interpreted by the Shapley additive exPlainations(SHAP)approach,which explicitly quantifies the effects of each feature in our model on the adiabatic temperature change and stress hysteresis.Additionally,we employ the Sureindependence screening and sparsifying operator(SISSO)method in conjunction with the key features to formulate explicit model.

关 键 词：Elastocaloric alloys Adiabatic temperature change Stress hysteresis Machine learning Multi-objective optimization 

分 类 号：TG132[一般工业技术—材料科学与工程]