Composition optimization of high strength and ductility ODS alloy based on machine learning

摘要

To find out characteristic parameters, mine data and predict candidate materials by machine learning is an important way for rapid development of materials. Based on about 300 groups data of composition, process, test condition and mechanical properties of ODS alloy, the correlation between the key component and the ultimate tensile strength and elongation of ODS alloy is established by deep learning method. It is found that there are optimal values of Cr%, Y2O3%, Al% and Ti% corresponding to the extreme value of the ultimate tensile strength. With the amount of Cr, Y2O3, and W increasing, the total elongation of ODS alloy decreased significantly, while the addition of Al and Ti is conducive to the improvement of the ductility in predicted range. Therefore, the optimized composition of higher strength and ductility ODS alloy is obtained through the correlation between the key components and tensile properties. The predicted tensile strength at room temperature is above 1400MPa, and more than 400MPa even at 700 degrees C. The predicted total elongation is more than 10 %. It will help to accelerate the optimization and development of ODS alloy which is used as structural material in fusion reactor.